Received August 10, 2019, accepted September 3, 2019, date of publication September 13, 2019,

date of current version September 26, 2019.

Digital Object Identifier 10.1109/ACCESS.2019.2940972

A Survey on Reversible Watermarking for

Multimedia Content: A Robustness Overview
ALEJANDRA MENENDEZ-ORTIZ1 , CLAUDIA FEREGRINO-URIBE2 ,
ROGELIO HASIMOTO-BELTRAN 1 , (Member, IEEE),
AND JOSE JUAN GARCIA-HERNANDEZ 3
1 Computer Science Department, Centro de Investigación en Matemáticas (CIMAT), Guanajuato 36023, Mexico
2 Computer Science Department, Instituto Nacional de Astrofisica, Optica y Electronica (INAOE), Puebla 72840, Mexico
3 CINVESTAV Unidad Tamaulipas, Ciudad Victoria 87130, Mexico
Corresponding author: Jose Juan Garcia-Hernandez (jjuan@tamps.cinvestav.mx)
This work was supported in part by the Mexican Council of Science and Technology (CONACyT)-FORDECYT-Consorcio de
Inteligencia Artificial under Project 296737, and in part by the Ph.D. Scholarship under Grant 351601, Grant PN-2017-01-5814, and
Grant PN-2017-01-7092 by CONACyT. Most of this work was developed while the A. Menendez-Ortiz was with the Instituto
Nacional de Astrofisica, Optica y Electronica in Mexico.

ABSTRACT In recent years, the growth of the Internet along with multimedia data delivery, have caused
the misuse and even alteration of the original digital content. As a security mechanism, digital watermarking
has been used to protect digital content, traditionally in copyright protection applications, and more recently
digital watermarking has been used in content enhancement applications. Regardless of the application,
digital watermarking schemes always cause some level of signal degradation. There are however, certain
applications as in the military or medical fields, where loss of information cannot be tolerated during the
reconstruction of host signals, therefore conventional watermarking schemes do not satisfy the quality
requirements. In order to reconstruct signals without loss of data, reversible watermarking is employed.
Recent research in reversible watermarking has focused on either, the improvement of embedding capacity
or the reduction of perceptual impact of the schemes, but scarcely exploring the robustness of reversible
schemes. In this survey, we provide a deep review of reversible watermarking schemes and propose a new
classification outline based on the robustness point of view. Finally, some future directions for reversible
watermarking research are presented and open problems in robust reversible watermarking are outlined.

INDEX TERMS Digital watermarking, multimedia content, reversible watermarking, robustness.

I. INTRODUCTION
Digital watermarking is the process of embedding additional
information, i.e. digital watermarks, into carrier signals.
These watermarks should, but do not need to be related to the
carrier signal. Digital watermarks are transparent or invisible
information patterns that are embedded into a signal, such
as images, audio or video [1]. Although visible watermarks
do exist, only invisible watermarks will be addressed in this
work. Invisible watermarks are perceptible only under certain
conditions, for example after applying their corresponding
detection algorithm, otherwise, they are invisible. Tradition-
ally, digital watermarking has been used in copyright pro-
tection, where the watermark contains information about the
copyright owner, and embedded in the digital content, in order
to prevent other parties from claiming the copyright [2]. More
recently, digital watermarking has been considered to enrich
The associate editor coordinating the review of this manuscript and to insert a large amount of information. The importance of
approving it for publication was Irene Amerini.
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see http://creativecommons.org/licenses/by/4.0/
132662
multimedia content by embedding meta-data as the digital
watermarks [3], [4].
Ideally, digital watermarking should not produce a notice-
able distortion between the original and the watermarked sig-
nal, and the watermarks should be difficult to remove without
damaging the host signal [5]. Depending on the application
where the watermarking scheme will be used, sometimes
the amount of information that can be embedded is also an
important characteristic. So, it can be seen that there is a
conflict between three properties in a watermarking scheme,
namely imperceptibility, robustness, and payload capacity.
A watermarking scheme that embeds a large amount of infor-
mation or inserts information in an imperceptible manner is
more susceptible to attacks, that is, some modifications to the
watermarked signal will cause the extraction to fail. On the
other hand, a scheme that resists a wide variety of attacks
with an acceptable degree of imperceptibility will not be able
these properties and the acceptable thresholds for each are
VOLUME 7, 2019
A. Menendez-Ortiz et al.: Survey on Reversible Watermarking for Multimedia Content: Robustness Overview
FIGURE 1. Elements in a conventional watermarking scheme.
determined by the applications where these schemes will
be used. For example, watermarking schemes for copyright
protection usually require more robustness than capacity and
schemes for multimedia enrichment are more concerned with
capacity rather than robustness.
A digital watermarking scheme, referred in this document
as conventional watermarking, consists on two processes:
embedding and extraction. In the embedding process a secret
message M is hidden in a host signal X , obtaining in that way
a watermarked signal Y that is transmitted over a communi-
cation channel where intentional or non intentional attacks
can occur. Therefore, the extraction process may receive a
watermarked signal Y or an attacked signal Ŷ . The extraction
process can have access to the host signal or not in order to
retrieve a message M 0 , also it may or may not retrieve a signal
X 0 . In the ideal case, where no attacks occur during trans-
mission, the recovered message M 0 is equal to the embedded
message M and the recovered signal X 0 is very similar to the
host signal X . Figure 1 shows the elements of a conventional
watermarking scheme.
A. EVALUATION METRICS
Two important properties of digital watermarking schemes
are the perceptual similarity between a host and a water-
marked signal, and the embedding capacity. In order to
measure the perceptual similarity of signals some metrics cal-
culate the differences between an original and a watermarked
signal. The peak signal to noise ratio (PSNR) [6] measures
the maximum signal to noise ratio found on a signal and
is typically used for image and video signals. The PSNR is
calculated with the following formula:
MAX2
PSNR(dB) = N × PN , (1)
n=1 (Xn − Yn )2
where N is the total number of samples in the signals, MAX
is the maximum possible value of a sample, Xn corresponds
to the n-th sample of the original signal X and Yn is the n-th
sample of the watermarked signal Y .
The signal to noise ratio (SNR) [6] reflects the distortion
that a watermark imposes over a signal and is typically used
for audio. The SNR is expressed in decibels and is given by:
PN
X2
SNR(dB) = 10 log10 PN n=1 n , (2)
2 carry payload suffer some level of degradation. Although this
n=1 (Xn − Yn )
VOLUME 7, 2019
TABLE 1. Ranges of ODG values.
where N is the total number of samples in the signals, Xn cor-
responds to the n-th sample of the original signal X and Yn is
the n-th sample of the watermarked signal Y . For both metrics
a higher value means that the watermarked signal is more
similar to the original one, i.e. it has a smaller degradation,
and a lower value means a bigger degradation. An objective
measure to evaluate audio quality is the objective difference
grade (ODG). The objective measurement algorithms model
the listening behavior of humans and their output is a num-
ber that describes the audibility of distortions. The objective
measurement method of the perceived audio quality (PEAQ)
is an international standard, ITU-R BS.1387, their values
range from 0 to −4 and are given in Table 1. This algorithm
compares the difference between a reference signal (original)
and a test signal (watermarked); both signals are processed
by an auditory system that calculates and estimates audible
components of the signal. These components can be consid-
ered as the representation of the signals in the human auditory
system. The internal representation is related to the masked
threshold, which in turn is based on a psychoacoustic model.
From these two internal representations an audible difference
is calculated and the cognitive model calculates the ODG
value from the audible difference [7].
As previously mentioned, another property of watermark-
ing schemes is the embedding capacity that refers to the
number of bits that a watermark encodes within a unit of time
or within a signal. The amount of bits that can be embedded
in an image are measured by bits per pixel (bpp) and the
bits that can be embedded within an audio are measured by
bits per second or by bits per sample (bps). The threshold
for acceptable degradation values, embedding capacity and
robustness depend on the applications where the watermark-
ing scheme will be used.
In conventional watermarking schemes, host signals that
132663
 A. Menendez-Ortiz et al.: Survey on Reversible Watermarking for Multimedia Content: Robustness Overview
FIGURE 2. Elements in a reversible watermarking scheme.
degradation can be controlled with a threshold, it is inevitable
that these modifications occur. The embedding process of
conventional schemes modifies the host signal in such a
manner that the original values are lost. The extraction pro-
cess recovers the hidden data and even though the retrieved
signal is very similar to the original one, it is not error free
because of the degradation introduced during the embedding
process. Even though this review focuses only on reversible
watermarking, there are works that overview robust strategies
for conventional watermarking, such as [8], [9].
Some applications require the use and transmission of
high-quality signals like in teleconferencing [10], navigat-
ing and alerting for the blind [11], warning and supporting
for traffic situations in vehicles [12], [13] or aviation [14]
applications that require high-quality audio; high-resolution
imagery applications include electronic presentations for
business, education and advertising, in the entertainment
industry for home theater or electronic cinema projections,
in military, transportation and medical applications, and game
simulation or training simulations [15]. In some cases, like
in the commercial field, it is desired that these high-quality
signals do not suffer loss of data. However, in the military and
medical fields, applications like treatment of voice disorders
or in systems for aviation warning, it is essential that these
signals do not suffer any kind of degradation.
Assume a medical application in tele-rehabilitation for
voice and speech disorders [16], where a doctor receives
an audio with a patient’s speech, analyses it and diagnoses
a treatment. A watermarking scheme can be used to insert
additional information about the patient’s current conditions,
so there is no need to send that data in an additional file.
If the additional data was embedded into the speech using
a conventional watermarking scheme, when the extraction
process retrieves the watermark there will be loss of data in
the recovered speech. In such scenario, a reliable diagnosis
cannot be made, because the doctor might have a speech with
a lower quality than the necessary to evaluate the voice or
speech of the patient. Even though conventional watermark-
ing schemes can be limited to low distortion with a quality
constraint, the host signal will not be completely recovered.
This example helps to illustrate that conventional water-
marking schemes are not sufficient for certain applications.
With this situation in mind, some modifications were made
to the conventional watermarking schemes and reversible
watermarking arose. Reversible watermarking schemes have
the advantage that the host signal can be reconstructed
after the watermarks are recovered, contrary to conventional
132664
watermarking where the host signal is lost and only a similar
version of the signal can be retrieved. Figure 2 shows the
elements in a reversible watermarking scheme.
In the same way as in conventional watermarking,
in reversible watermarking a secret message M is hidden into
a host signal X , resulting in a watermarked signal Y and
is then transmitted through a communication channel. The
extraction process requires Y and control data in order recon-
struct the host signal X and it also recovers a message M 0 .
The difference with conventional schemes is their ability to
perfectly reconstruct the host signal in a lossless manner, but
to do so, control data is essential.
In this survey, we focus on analyzing the capacity
of reversible watermarking schemes to resist attacks, i.e.
their robustness. For many years, reversible watermarking
schemes were intrinsically fragile, i.e. the schemes did not
deal with any kind of attacks at all; these schemes worked in
scenarios that supposed a noise free communication channel
for transmitting the watermarked signals. However, as sce-
narios grew more elaborate, attacks began to be considered
as an inevitable condition in the communication channel, and
therefore these considerations had to be included in the design
of reversible watermarking.
As it was previously mentioned, this review focuses
on reversible watermarking schemes and more precisely
on the robustness of reversible strategies. The classifica-
tion proposed in this manuscript is presented in Figure 3.
Reversible watermarking schemes are fragile by nature,
a review of these techniques is given in Section II, compre-
hensive reviews on fragile reversible watermarking schemes
for images have been reported in [17], [18]. The rest of
the categories from Figure 3 are reviewed in Section III
of this manuscript. Robust reversible watermarking can be
further divided in semi-fragile reversible watermarking which
resists fewer attacks than robust reversible watermarking
schemes. Although self-recovery watermarking schemes can-
not be considered reversible watermarking schemes in a strict
sense, we include them in Section III because they present
a sort of robustness in which they allow the restoration
of the watermarked signals, although they do not carry a
secret message within. Self-recovery watermarking schemes
can be further classified as those with perfect and approxi-
mate reconstruction. Finally, a special case of robustness that
has been recently studied is also analyzed in this section,
namely reversible watermarking with watermark and signal
robustness, this special kind of scheme combines strate-
gies from fragile reversible watermarking and self-recovery
VOLUME 7, 2019
A. Menendez-Ortiz et al.: Survey on Reversible Watermarking for Multimedia Content: Robustness Overview
FIGURE 3. Proposed classification for reversible watermarking schemes.
watermarking to design schemes that are capable of both
extracting the watermark information and reconstructing host
signals.
The rest of the document is organized as follows: a
review of the most significant fragile reversible watermarking
schemes is presented in Section II. A review on robustness in
reversible watermarking schemes is presented in Section III.
The findings from this extensive review are summarized in
Section IV. Finally, the conclusions and open problems of the
work are given in Section V.
II. FRAGILE REVERSIBLE WATERMARKING SCHEMES
In the previous section, it was stated that conventional water-
marking schemes are not a suitable option for applications
that require intact signals after watermark extraction, i.e.
that require signals that maintain very high quality. For this
reason, reversible watermarking schemes were proposed. The
latter schemes insert watermarks and after extraction the
modifications caused by embedding are reversed in order to
produce non-altered signals. However, because the reversibil-
ity requirement imposes a big restriction, most reversible
watermarking schemes in the literature are fragile by nature,
i.e. their ability to extract the watermark and restore the origi-
nal values of the signal is restricted only to those watermarked
signals that were not altered in any manner. In this section, a
review of the most relevant fragile reversible watermarking
techniques used for images and audio signals is presented.
VOLUME 7, 2019
These schemes are reviewed by the watermarking technique
they use, namely error expansion based and histogram shift-
ing based schemes. In this section, comparisons of schemes
are presented and the results they report are detailed; BPP is
used to measure the payload they insert and PSNR is used to
measure the degradation of the watermarked images; BPS is
used to measure payload, SNR and ODG are used to measure
the quality of watermarked audio signals. Although other
metrics exist for both images and audio signals, these are
the only ones adopted in this review because they are the
most commonly used metrics throughout the literature, other
metrics are not adopted by all the works and a comparison of
these metrics is not possible.
A. ERROR EXPANSION BASED SCHEMES
Error expansion is the technique most used in reversible
watermarking, both for images and audio signals. The general
idea is to calculate an error, that can be obtained directly
from a difference between pixels or samples, or it can be
the error between the sample and its predicted value, for
instance. The error is expanded and the watermark bit is
inserted in the least significant bit (LSB). The most relevant
techniques of error expansion are Difference Expansion (DE),
Interpolation-Error Expansion (IEE), and Prediction-Error
Expansion (PEE). In [17] a review on image reversible water-
marking schemes is presented, with special focus on predic-
tion error expansion techniques. In this section we elaborate
132665
 A. Menendez-Ortiz et al.: Survey on Reversible Watermarking for Multimedia Content: Robustness Overview
TABLE 2. Most relevant error expansion schemes for images.
TABLE 3. Most relevant error expansion schemes for audio.
on both image and audio reversible watermarking schemes
based on error expansion. The most relevant schemes for
image reversible watermarking are presented in Table 2, and
schemes for audio signals are presented in Table 3.
1) AMPLITUDE EXPANSION
Amplitude expansion technique has only been proposed for
audio reversible watermarking schemes, the idea behind it
is to simply expand the time-domain values of the sam-
ples to make room for watermark insertion in the LSB.
A companding strategy is introduced to control the degra-
dation introduced by this expansion. The scheme by [32]
is the first to propose this strategy, where the value of one
sample, i.e. its amplitude, is expanded and a watermark
bit can be inserted in the LSB. To control underflow and
overflow problems, a companding (compression and expan-
sion) approach is introduced, where quantized compression
and quantized expansion functions are proposed. The audio
signal is compressed, then a left bit-shift is performed so
the watermark can be embedded; the compressed signal is
expanded and a companding error is obtained; the compand-
ing error is compressed and forms part of the watermark.
The watermark consists of the data to be embedded, the com-
pressed companding error and overhead data. In [33] a fragile
steganography scheme for integrity verification applications
is presented, it uses amplitude expansion in intDCT domain.
132666
IntDCT coefficients are grouped into M regions, the first
M /2 groups are reserved for amplitude optimization and
remaining groups are used for watermark embedding. Posi-
tional data indicate groups that were embedded and groups
where amplitude was optimized to avoid overflow. After
embedding, the inverse intDCT transform is applied to obtain
the watermarked audio. In [34] a scheme is presented where
two watermarks are embedded: a robust and a reversible
watermark. Under no attacks, the scheme can recover both
watermarks. The robust watermark can be extracted after
bandpass filtering, Gaussian noise, and additive delay. The
scheme uses intMDCT coefficients to insert two watermarks.
The robust one is embedded using spread-spectrum modula-
tion based on a pseudo-random number (PRN) sequence. The
reversible watermark is embedded in frequency coefficients
as well, using an amplitude expansion technique.
2) DIFFERENCE EXPANSION
Difference expansion technique calculates a difference
between current pixel or current sample and another pixel
or sample from its neighbor area. This difference is then
expanded to make space for the watermark bit. In [19] a
combined strategy for insertion in spatial prediction error and
frequency prediction errors is proposed. The first stage inserts
watermark bits in spatial domain using PEE, and the second
stage in high frequency obtained through an integer Haar
VOLUME 7, 2019
A. Menendez-Ortiz et al.: Survey on Reversible Watermarking for Multimedia Content: Robustness Overview
wavelet transform using DE. In [20] a scheme is proposed
where a difference is calculated using four pixels, and it is an
improvement to Tian, et al. scheme [44]. In [21] their previous
transformations are generalized in order to insert watermark
bits in three or four differences to N differences. They use
each color component to carry payload, therefore increasing
embedding capacity. In [22], differences are calculated on
overlapped blocks, and watermark bits can be embedded in
the differences of these overlapped blocks, i.e., the theoretical
embedding capacity is 1 bpp. The scheme can be applied
multiple times (multi-embedding) achieving an embedding
capacity of 2 bpp with acceptable degradation. However,
practical limitations of multi-embedding capacity vs. degra-
dation have to be further explored. A scheme for audio is
proposed by [35], it is based on a generalized reversible
integer transform (GRIT) originally proposed for images. For
each audio frame, samples of close proximity are grouped
into vectors of size 2 and 3 (for dyad and triplet transforms).
Experimental results are obtained with triplet and dyad-based
transforms, where the dyad-based transform is identical to
Tian’s scheme. This work shows that a difference expansion
technique can be applied to audio signals and it was found
that the triplet-based transform achieves higher embedding
capacities than the dyad-based one with similar peak signal-
to-noise ratio (PSNR) results.
3) INTERPOLATION ERROR EXPANSION
This technique is based on the idea that by interpolating the
values of neighbor pixels or samples, a better estimation can
be obtained and therefore the error can be reduced. In [23]
a Particle Swarm Optimization (PSO) strategy is used to
determine the optimal weights used to interpolate the cur-
rent pixel estimation. In [24] a new rhombus interpolation
is proposed, in a simple and straightforward manner they
determine if the prediction is to be obtained with four or
two neighbors, and also determine if a horizontal or vertical
prediction is more appropriate. After this prediction, classical
PEE is applied, histogram shifting is used to reduce the size
of the location map, and a compressed location map along
with a header are inserted in the reserved pixels. Experimental
results demonstrate that the proposed interpolation prediction
outperforms MED and GAP predictors. In [25] a genetic
algorithm is used to determine weights of neighboring pixels
used for interpolation prediction. The scheme by [36] is based
on classical interpolation error expansion for images, with an
adaptation of processing frames of samples for audio signals.
In this work, a new methodology to construct a boundary
map is proposed, where a quantization factor is considered.
Experimental results reported in this work indicate how the
quantization bits affect the embedding capacity, the signal
degradation and the size of the boundary map. With more
quantization bits there is lower embedding capacity, lower
degradation and the boundary map is smaller; with less quan-
tization bits there is higher embedding capacity, higher degra-
dation and the boundary map is bigger; however, by reducing
quantization bits, embedding capacity and the boundary map
VOLUME 7, 2019
do not increase in the same ratio, i.e. the size of the boundary
map becomes a minor concern given the embedding capacity
obtained.
4) PREDICTION ERROR EXPANSION
This technique explores different sorts of strategies to obtain
predictors, which can reduce the prediction error and there-
fore reducing the degradation. The idea that [26] introduces is
embedding the watermark bit on the context pixels in order to
reduce the perceptual impact of the embedding process. The
theoretical construction is tested using three different predic-
tors, namely MED, GAP and SGAP. These implementations
are also optimized to further improve transparency. In [27] a
modification to classical PEE is proposed and two thresholds
are introduced to control capacity and degradation of the
scheme. The scheme by [28] presents a multi-bit embed-
ding strategy that modifies classical PEE, where instead of
expanding the prediction error two times, this prediction error
is expanded n times to allow the insertion of more than
one bit into the LSBs. A combined strategy of Histogram
Shifting and overflow/underflow map is used at detection to
identify which pixels do not carry watermark information.
Histogram shifting modifies the pixels so their values are
greater than a fixed threshold, at detection pixels that exceed
the threshold are re-adjusted to their original value without
extracting information. The remaining pixels that cannot be
shifted are marked in an underflow/overflow map, which has
better compression than a classical location map. In [29]
they continue their previous work of using two thresholds for
controlling the embedding of the watermark, but proposing
improvements to reduce the embedding degradation and yet
maintaining the fine tuning for embedding capacity. In [30]
a local prediction-based scheme is proposed, as opposed to
classical schemes based on global prediction. For each pixel,
a block of size B × B is selected and a local predictor is
derived for each pixel calculated with least squares (LS).
At detection, the same predictors can be derived by extracting
in reverse order and by substituting the central pixel with an
approximation (obtained with a classical fixed predictor such
as MED, GAP, SGAP). Pixels in the borders of the image are
calculated with fixed classical predictors as opposed to the
local predictors. A histogram shifting strategy is implemented
for pixels that do not carry watermark information, and a
location map is constructed to mark underflow and overflow
pixels. This local prediction-based scheme outperforms other
schemes in the literature, at the cost of greater computational
cost. In [31] an adaptive pixel pairing is presented to reduce
distortion, considering low embedding capacities. They use
error expansion to insert the watermark bits in every pair of
pixels and a histogram shifting strategy to reduce the pixels
that need to be marked in the location map.
The work in [37] presents a prediction error expansion
scheme for audio signals. In this work, an integer coeffi-
cient prediction is obtained from consecutive audio samples
weighted by three prediction coefficients; these coefficient
predictors are obtained from a coefficient matrix that contains
132667
 A. Menendez-Ortiz et al.: Survey on Reversible Watermarking for Multimedia Content: Robustness Overview
all the permutations of the weights. They use a location map
instead of a histogram shifting strategy to control underflow
and overflow. The location map is compressed and appended
to the watermark along with the prediction coefficients. The
extraction process uses the prediction coefficients and the
decompressed location map to extract the watermark and
restore the audio samples. A threshold T can control the
embedding distortion caused by the scheme, however it also
limits the embedding capacity. In [38] a reversible audio
watermarking scheme is proposed and it uses the general
idea of prediction error expansion from [45] but adapted for
audio signals. A n-th order autoregressive linear prediction
is proposed and adaptive weights are used to obtain a linear
prediction with a Burg method [46]. In [39] another scheme is
proposed and is based on a previous scheme. This work uses
a modified prediction error expansion technique to insert a
watermark. In the strategy presented, the prediction values
and prediction errors are calculated in the same way as in
the previous work, but an expansion factor β, where β > 1,
is introduced to control the expansion factor, for β = 2
the method is the same as the previous one. A rounding
function F(.) is also introduced, where the prediction error
is first compressed by β and then rounded. The watermark
bit is embedded if the prediction error can be recovered
through the compression and rounding performed by F(.).
In [40] a prediction error expansion scheme based on Thodi
and Rodriguez’ original method [45] is proposed. In this
work, a non-causal predictor is proposed to be used for
reversible audio watermarking. This predictor presents better
performance than other predictors in the literature. The linear
non-causal predictor proposed by the authors takes past and
future values of the signal to calculate a prediction value.
Embedding of the watermark is performed in two stages, for
each stage two groups of samples are selected; the first group
consists of odd samples and the second of even samples. Two
samples of the first group are used to calculate one prediction
value for the second group; the prediction errors are obtained
by subtracting the prediction values from the original sample
values of the second group. A second stage of embedding
takes the watermarked samples in the second group to obtain
the prediction values of the first group; the prediction errors
are obtained subtracting the prediction values from the water-
marked sample values of the first group. The scheme by [41]
proposes a new method to calculate non-causal predictors
that reduce prediction-errors in an adaptive way. Minimum
error power is used to find the minimum prediction-errors
for the selected context and prediction coefficients. A double
embedding approach is adopted. It divides audio samples
in two sets, the prediction-context is adaptively selected for
the first set and watermark bits are embedded with those
parameters, later the prediction-context of the second set is
found using the modified samples from the first set. How-
ever, in order to find adequate prediction coefficients and to
find the minimum prediction-errors, two parameters must be
found, namely K1 and K2 . The performance of the whole
strategy depends on the values of these two parameters, but
132668
TABLE 4. Most relevant reversible histogram shifting schemes for images.
their calculation is not clearly described. In [42] a strategy
originally proposed for images is adapted to classical audio
PEE. The main idea is to insert the watermark bits using two
thresholds, namely T and T − 1, for capacity control. First D
bits are inserted using a threshold T , then the remaining bits
are inserted using T − 1. The two-threshold strategy can be
implemented independently of the prediction used. Optimal
results are obtained for 0.6 bps. In [43] two reversible water-
marking schemes for audio authentication are presented, one
uses difference expansion based on Tian’s work [44] and
the other uses prediction error expansion based on Thodi
and Rodriguez’ work [45]. Both the difference expansion
method and the prediction error expansion method are very
similar to the original works for images, except that they
work with the integer discrete cosine transform (intDCT)
coefficients obtained from the host audio signals. The original
difference expansion method uses consecutive pixel pairs to
calculate the differences to be expanded, whereas the audio
implementation uses past and future odd intDCT coefficients.
B. HISTOGRAM SHIFTING BASED SCHEMES
Another technique for reversible watermarking broadly used
is histogram shifting. The main idea of histogram shifting is to
obtain the histogram of gray level pixels in the image, then the
highest and lowest bin in the histogram are found, the lowest
bin is emptied and the positions of the pixels corresponding
to that gray level intensity are saved for reversibility, then
the bins between the highest and lowest bin are right shifted
by adding one to the gray level values of the corresponding
pixels; watermark insertion is performed on those pixels that
correspond to the new empty bin. Table 4 presents some
of the most relevant works that propose histogram shifting
strategies.
In [57] a scheme with 48.5 dB is reported, it uses his-
togram shifting and its contribution is the introduction of a
down-sampling of the images prior embedding, however an
up-sampling is required after data insertion; the location map
constructed during embedding is used to detect tampering
and collage attacks. In [54] 59 dB of PSNR are reported
for a scheme that uses histogram shifting and do not require
additional information to restore the host image, it uses
a shifting strategy to control the underflow and overflow
VOLUME 7, 2019
A. Menendez-Ortiz et al.: Survey on Reversible Watermarking for Multimedia Content: Robustness Overview
problem, thus it does not require a location map. The
scheme is designed for images in Digital Imaging and
Communication in Medicine (DICOM) format. Nonetheless,
the embedding strategy uses many cases that can increase the
complexity of the algorithm and those cases depend on the
size of the blocks that are used for processing the images.
In [53] a work that reaches 56.7 dB is presented and it uses
dynamic histogram shifting with two types of modulation,
pixel histogram shifting and dynamic prediction-error his-
togram shifting; it can insert data in areas of the image with
texture, where other schemes fail and it has an acceptable
trade-off between embedding capacity and distortion. In [55]
a scheme is proposed and it reports the highest embedding
capacity found to date with 6.56 bpp, it also has a low
perceptual impact with a PSNR value of 54.3 dB; this scheme
uses a histogram shifted imitated technique that can use a
multi-layer embedding and allows a higher embedding capac-
ity. The results reported were obtained with 10 layers, but
with one layer it only reaches 0.5 bpp. This version of the
histogram shifting technique does not reach a high embed-
ding capacity, but instead of using as reference the highest
peak of the histogram from the whole image, the histogram
is taken from segments of the image. The perceptual impact
depends on the segmentation of the image and can be set
very low, so a multi-embedding technique can be used to
increase embedding capacity without significant distortion to
the watermarked image. In [56] a scheme is proposed and
it obtains 48.7 dB of PSNR and it uses histogram shifting;
interpolation errors are calculated between two sets of pixels
previously selected and a histogram is constructed from those
errors, data is embedded by shifting the histogram bins; this
scheme can improve its embedding capacity if it is applied
multiple times. The scheme by [50] reaches 51.8 dB of
PSNR by using histogram shifting, the histogram is con-
structed from the differences between the original image
and a difference image; by applying the scheme multiple
times, it can reach 1.4 bpp and 36 dB, however the control
information grows with each iteration. In [49] a scheme is
presented, it uses histogram shifting and reports a PSNR
value of 48.3 dB; the embedding is performed by modify-
ing the pixels based on their tonal distribution, it does not
require a location map in order to solve the underflow and
overflow problem, but its embedding capacity is low. The
scheme by [51] reaches 47 dB and uses histogram shifting,
which is constructed by calculating the differences between
neighboring pixels instead of the direct values of the pixels;
although its embedding capacity with one layer is small it
could be improved by applying the scheme multiple times.
The work by [52] embeds 1.16 bpp using a combined tech-
nique of histogram shifting and error prediction, the errors
are calculated with rhombus prediction and the histogram
is constructed from those errors, the embedding capacity
is reached by using multiple bins to insert the watermark.
In [48] a payload of 1.51 bpp is inserted, the scheme uses
histogram shifting and allows a double stage insertion with
low perceptual impact; however, the double insertion cannot
VOLUME 7, 2019
be performed in all the bins from the histogram. The scheme
by [47] obtains a PSNR value of 49.4 dB and uses histogram
shifting, it calculates differences among pixels and constructs
a histogram with two peaks from those differences, each peak
is used as a reference for embedding; the scheme is fragile,
it can detect tampering and it is used for authentication.
As it can be seen, prediction error expansion is the most
explored technique in the literature for both images and
audio signals. Since fragile reversible watermarking schemes
assume scenarios without attacks they do not concern with
robustness and instead they try to increase embedding capac-
ity and to maintain acceptable levels of transparency. Fewer
strategies have been proposed for audio signals than images,
although some prediction error expansion schemes present
competitive results.
Next section presents an analysis of reversible watermark-
ing schemes that consider robustness as a desirable property.
A classification in terms of the type of robustness presented
by these schemes is proposed.
III. REVIEW OF ROBUSTNESS IN REVERSIBLE
WATERMARKING
Reversible watermarking schemes can exhibit a certain
degree of robustness against attacks, i.e., their capacity to
extract watermarks after some modifications have occurred
to the watermarked signal. This robustness property has been
rarely explored and so far, robustness in reversible water-
marking has been approached from the following points of
view: robustness of the watermark, and a special type of
schemes called self-recovery that allow signal restoration.
Although self-recovery schemes cannot be strictly catego-
rized as reversible watermarking schemes, we include them
in this section because of their capacity to counteract modifi-
cations made by attacks and to regenerate original signals.
A. ROBUST REVERSIBLE WATERMARKING
Robust reversible watermarking schemes allow the recovery
of watermarks and the host signal when attacks are not present
in the communication channel. When attacks do occur, only
the watermarks can be extracted, and there is no guaran-
tee that the host signal can be reconstructed. In an attack
situation, the recovered signal may be very similar to the
host one, nonetheless it is not exactly the same, thus in this
situation there is no reversibility. These two scenarios are
depicted in Figure 4. Based on the watermark robustness,
these schemes can be further classified as semi-fragile or
robust.
1) SEMI-FRAGILE WATERMARKS
Reversible schemes with semi-fragile watermarks have a cer-
tain degree of robustness against a reduced set of attacks.
The watermarks in the semi-fragile schemes resist only unin-
tentional attacks, like slight compression; on the other hand,
the watermarks in the robust schemes should be able to
survive intentional attacks, like signal processing operations.
The former were the first efforts to preserve the embedded
132669
 A. Menendez-Ortiz et al.: Survey on Reversible Watermarking for Multimedia Content: Robustness Overview

FIGURE 4. Elements in a robust reversible watermarking scheme.

watermarks after some kind of processing was applied to the TABLE 5. Reversible watermarking schemes with semi-fragile
watermarks.
watermarked signal.
The first semi-fragile reversible watermarking scheme
found in the literature is the one proposed by [58], it can detect
regions of the image that were tampered but the watermarked
images suffer from salt-and-pepper noise. In [59] a scheme is
proposed and it works with images in the spatial domain and
resists JPEG compression; however, the watermarked images
suffer from salt-and-pepper noise. In [60] and [61] the prob-
lem of salt-and-pepper from the previous schemes is solved;
this scheme also works with images in the spatial domain and
resists JPEG and JPEG2000 compression; nonetheless, this
scheme has low embedding capacity. In [62] and [63] another
scheme that solves salt-and-pepper is presented, images in
the integer wavelet transform (IWT) domain are used and a robust (semi-fragile) and a fragile one. The robust
the scheme resists JPEG2000 compression, but the embed- watermark is inserted in mid-low frequencies in IWT, and
ding capacity is also low. In [64] a scheme is proposed, the fragile watermark contains information for reversibility,
it uses IWT domain and resists JPEG compression. In [65] the fragile watermark is inserted in spatial domain using
a scheme is designed where images embedded in the spatial histogram shifting. The scheme is robust agains JPEG com-
domain resist JPEG compression. The work by [66] proposes pression. Table 5 depicts a summary of these schemes.
a semi-fragile scheme based on IWT transform and histogram
shifting technique, the cover image is divided into non- 2) ROBUST WATERMARKS
overlapping blocks, IWT coefficients are obtained for each The first robust reversible watermarking schemes appeared
block and then a multi-level histogram shifting mechanism in 2007 with the work by [69] where a scheme that works
is applied to insert the watermark bits. When no attacks with images in the spatial domain is proposed, it is robust to
occur the schee is completely reversible and it is semi-fragile various attacks such as flipping, rotation, up-sizing, increas-
against JPEG compression. In [67] a semi-fragile reversible ing aspect ratio, cropping, drawing, among others; however,
watermarking scheme is proposed and based on Slantlet the embedding capacity is low. In [70] a scheme that works
transform (SLT). The cover image is divided into blocks and in integer transform domain (ITD) is presented and is robust
SLT coefficients of each block are obtained for watermark against cropping. The work by [71] proposes a scheme that
insertion. This scheme is robust against JPEG compression. works with magnetic resonance images in spatial domain and
In [68] a semi-fragile reversible watermarking scheme is pro- it is robust against JPEG compression. In [72] a scheme that
posed, it is based on IWT, and two watermarks are inserted: works in IWT domain is introduced and it has robustness

132670 VOLUME 7, 2019

A. Menendez-Ortiz et al.: Survey on Reversible Watermarking for Multimedia Content: Robustness Overview
TABLE 6. Reversible watermarking schemes with robust watermarks.
to cropping and salt-and-pepper noise. In [73] a scheme for
images and signals in the spatial and temporal domain is
proposed, both images and signals can be processed in the
transform domain. This scheme is robust against the addition
of white Gaussian noise (AWGN). The work by [74] presents
a scheme in wavelet domain that is robust against JPEG com-
pression. In [75] a scheme that works with images in discrete
cosine transform (DCT) domain is introduced and it is robust
to blurring, brightness, contrast and cropping, among others.
In [76] a method in IWT domain is proposed and it presents
robustness against brightness, JPEG and JPEG2000 com-
pression, cropping and inversion. In [77] a scheme in DWT
domain is presented that is robust to JPEG compression,
AWGN, salt and pepper noise, scaling and blurring, among
others. In [78] a scheme that uses images in spatial domain
is introduced and it is robust against JPEG compression.
The work by [79] also presents a scheme that works with
images in spatial domain and is robust to JPEG compression.
In [80] a scheme that works with images in IWT domain is
designed and it is robust against brightness, contrast, pos-
terization, inversion, cropping, JPEG, and JPEG2000 com-
pression. In [81] and [82] schemes that work with images in
spatial domain are presented, both are robust against JPEG
and JPEG2000 compression, and additive AWGN. In [83] a
scheme that uses images in wavelet domain is introduced,
it is robust against JPEG and JPEG2000 compression, and
additive AWGN as well. The work by [84] presents a robust
reversible watermarking scheme that uses a logistic chaotic
map to select the regions for data hiding and also con-
trols a threshold that is used to balance reversibility and
robustness. Because of the characteristics of chaotic maps,
the scheme also has an increased security compared against
VOLUME 7, 2019
other schemes in the literature. In [85] a scheme is proposed,
based on the Slantlet transform (SLT). The image is divided
into non overlapping blocks, each block is transformed using
SLT, and the mean value of a chosen high frequency sub-
band is modified to carry watermark information. Overflow
and underflow are controlled with a histogram shifting tech-
nique. This scheme is robust against JPEG, JPEG2000, addi-
tive Gaussian noise (AGN), and salt and pepper noise, and
robust to some extent against sharpenning and histogram
equalization. The work by [86] proposes a robust scheme
for color images, it is based on pixel-value-offset of each
block from the image. Table 6 presents a summary of these
schemes.
B. SELF-RECOVERY WATERMARKING
Although self-recovery schemes are not reversible water-
marking schemes in the strict sense, they are worth to be
included in this overview because their capacity to recon-
struct original signals, whether attacks occurred in the trans-
mission channel or not. This is an important aspect to notice,
because reversible watermarking lacks of this characteristic;
i.e. in the presence of attacks, robust reversible watermarking
schemes recover the watermarks but not the host signal,
and self-recovery schemes recover that host signal. Nonethe-
less, self-recovery schemes cannot embed useful payload.
In other words, their whole embedding capacity is used to
insert control data and no watermarks at all. Control data
contain either a compressed version of the signal or signif-
icant characteristics that help in the reconstruction process.
Self-recovery schemes can be further categorized as schemes
with approximate or perfect reconstruction and are explained
below.
132671
 A. Menendez-Ortiz et al.: Survey on Reversible Watermarking for Multimedia Content: Robustness Overview
FIGURE 5. Elements in a self-recovery scheme with approximate reconstruction.
FIGURE 6. Elements in a self-recovery scheme with perfect reconstruction.
1) APPROXIMATE RECONSTRUCTION
Figure 5 shows the elements in a self-recovery scheme with
approximate reconstruction. This type of schemes can detect
tampered regions of a signal and try to reconstruct the tam-
pered regions using the embedded control data. Although
this reconstructed signal may be similar to the original sig-
nal, it is not a lossless version of the host signal. Table 9
presents the self-recovery schemes with approximate signal
reconstruction. The first works that recovered a signal after
this was subjected to attacks were proposed by [87], [88].
From these works many others have been proposed, some of
them modify the domain where the embedding takes place,
others improve the compression method to calculate control
data. Some relevant works will be briefly described below.
As previously mentioned, the first works to reconstruct
an image after some malicious operation was applied to the
watermarked signals were [87], [88]. These schemes utilized
DCT domain to embed a compressed version of the image
into itself. In [89] the first scheme for video signals was
proposed and it was a modified version of the MPEG (Moving
Picture Experts Group) compression standard. In [90] the
first scheme for audio signals was proposed, it inserts the
watermarks in temporal domain and the scheme was only
able to detect the places where the audio had been modified,
but could not correct these modifications. In [91] the first
scheme for audio signals able to correct modifications by
cropping was proposed, with a certain degree of accuracy
for correction. Recent works like the ones proposed by [92]
and [93] improve the quality of reconstructed images after
they have been subjected to attacks. But even though these
132672
are high-quality images, they are not lossless versions of
the original image. In [94] a self-recovery scheme based on
singular value decomposition (SVD) is proposed. The cover
image is divided into blocks, and from each block its singular
values are obtained, two detection keys are generated from
those values, these keys are unique for each block. A random-
block mapping sequence and optimization strategies are used
for tamper detection.
2) PERFECT RECONSTRUCTION
Figure 6 shows the elements in a self-recovery scheme with
perfect reconstruction. Like the schemes with approximate
reconstruction, these can detect the regions where the signal
was tampered and with the embedded control data, they can
reconstruct the tampered regions. However, the reconstructed
signal has exactly the same values that the ones in the host
signal, i.e., they achieve perfect signal reconstruction. Table 7
presents the self-recovery schemes with perfect signal recon-
struction capability. All of them were proposed for image
signals and work in the spatial domain. The first method was
proposed by [95], where the image is perfectly reconstructed
after content replacement, as long as tampered areas are not
too extensive. In [96] the scheme from [95] is improved,
it can also reconstruct the image after content replacement
attack as long as the tampered area is smaller than 24% of
the image. In [97] a scheme that can perfectly reconstruct
an image after cropping attack is presented, as long as the
tampered area is smaller than 25% of the image. In [98]
their previous approach from [97] is improved in terms of
computational cost for tamper restoration. This method, as
VOLUME 7, 2019
A. Menendez-Ortiz et al.: Survey on Reversible Watermarking for Multimedia Content: Robustness Overview
TABLE 7. Self-recovery schemes with perfect reconstruction.
FIGURE 7. Elements in a reversible watermarking scheme with watermark and signal robustness.
well as the previous one performs the restoration of tampered
areas in an iterative manner, where the first iterations yield
approximate restoration, but as the iterations increase so does
the restoration quality. In [99] a self-recovery scheme for
audio signals is presented. To the best of our knowledge, this
is the first effort in the literature to propose such a scheme for
audio signals. A reduced version of the audio samples in time
domain, i.e. reference bits, is calculated, and hash bits are
obtained to identify tampered segments in the received audio
signals. Both reference bits and hash bits are inserted into
intDCT high frequencies using a reversible watermarking
technique, namely prediction-error expansion.
The drawback of both types of self-recovery schemes is
that they can reconstruct the signal only if the tampered area
is not too extensive. If too many regions of the signal were
corrupted, then the values of the compressed version that
was originally embedded may be corrupted or lost during
attacks. Because of this, not all the tampered regions may be
reconstructed and if there were too many tampered regions,
the whole image would be lost.
From the robust reversible watermarking schemes and
self-recovery schemes analyzed so far, an interesting relation-
ship arises. Robust reversible watermarking and self-recovery
schemes seem to complement each other in a general objec-
tive: recover what was lost during attacks. Robust reversible
watermarking recovers the watermarks, while self-recovery
recovers the host signal. Considering that robust reversible
watermarking and self-recovery schemes can complement
each other, a new type of scheme was proposed, exploring the
idea of having a reversible watermarking scheme that, under
certain circumstances, can reconstruct both the watermark
and the host signal in the presence of attacks.
C. REVERSIBLE WATERMARKING WITH WATERMARK
AND SIGNAL ROBUSTNESS
In this scenario, a watermarking scheme can reconstruct the
host signal and the watermarks, regardless if certain attacks
VOLUME 7, 2019
were present during transmission, as depicted in Figure 7.
This problem was originally addressed from the point of view
of communications theory. It is possible to reconstruct the
watermark and the host signal after some theoretical attack
within a symmetric binary transmission channel. However,
even if some specific attack is established, this kind of scheme
is not suitable for practical watermarking applications.
Watermark and signal robustness were first analyzed
by [100], where a theoretical analysis of this situation is
given. In this work a scenario is presented where a binary
mark is embedded in a binary host signal, which in turn
is transmitted through a symmetric binary channel. This
transmission channel can cause a certain distortion to the
watermarked signal and/or the embedded data. A host sig-
nal (image) is compressed in a lossless manner and a
bit-stream of auxiliary data is concatenated. This auxiliary
data contains the watermark and control data to reconstruct
the host image. This scheme is extended to work in channels
with attacks and it is solved by attaching error correcting
codes to the auxiliary data. In this paper, only the theoret-
ical foundations are given for an over-simplification of the
problem, and although this is a valid scenario for a commu-
nications problem, it does not satisfy the idea behind digital
watermarking where the watermarked signal must appear
unaltered to the unaware observer.
The works by [101]–[108], and [109] continue the theoreti-
cal analysis given by [100], trying to improve the compressor,
thus achieving higher theoretical payload capacities. In [110]
a code construction that enhances payload capacity is pro-
posed and an example on how to improve a histogram shifting
method by including these codes is explained. This modi-
fied histogram shifting method only supports binary images.
In [111] an extension to their codes to gray-scale signals is
presented and two histogram shifting methods are modified
to include codes for gray-scale images. These modifications
enhanced payload capacity given a distortion constraint; how-
ever, only the distortion caused by the embedding process
132673
 A. Menendez-Ortiz et al.: Survey on Reversible Watermarking for Multimedia Content: Robustness Overview

FIGURE 8. Chronology of the watermark and signal robustness works from communications theory.

TABLE 8. Reversible watermarking schemes with watermark and signal robustness.

is considered, not the distortion caused by attacks. Figure 8
depicts a summary of these works and their most important
characteristics.
Some of the works analyzed so far have tackled the prob-
lem when malicious attacks occur in the transmission channel
by converting the non-binary host signal into a binary host
signal and then applying existing methods. However, that is
not the objective of digital watermarking for multimedia con-
tent. Because of that, some efforts have proposed reversible
watermarking schemes that, to some extent, can restore the
watermark and the host signal after certain attacks occur
during transmission.
In order to achieve both types of robustness, a framework
was proposed by [112] where a fragile reversible watermark-
ing scheme is concatenated with a self-recovery scheme with
perfect restoration. After attacks, self-recovery restoration is
applied first to restore the host signal in a lossless manner,
and later the watermark is extracted with the fragile reversible
scheme. In this way, two independent schemes are combined
in a framework to be used as one scheme.
Table 8 presents the works that have designed reversible
watermarking schemes with watermark and signal robust-
ness. The first to introduce the idea was proposed by [112]
where an approach for images was presented. Later, a mas-
ter’s thesis [113] follows the construction proposed by [112]
and implements several combinations of fragile reversible
watermarking schemes and three self-recovery schemes for
images with perfect restoration, presented in Table 7. Later,
a PhD thesis implements this kind of framework for audio
signals; a reversible watermarking scheme for audio [115]
and a self-recovery scheme for audio [99] had to be designed
to complete the framework.
In summary, robust reversible watermarking schemes
have proposed solutions that restore the watermarks, and
self-recovery schemes recover the host signal when the
watermarked signal suffers modifications caused by attacks.
Robust reversible watermarking schemes can be semi-fragile,
as the works listed in Table 5, and more recent efforts have
presented reversible schemes with robust watermarks as the
works given in Table 6. Self-recovery schemes can have
approximate reconstruction of the signal, like the works pre-
sented in Table 9; but very few have perfect reconstruc-
tion of the signal, only five have this property and are
given in Table 7. Recent works have proposed a new type
of reversible watermarking scheme that combines fragile
reversible watermarking with self-recovery schemes with

132674 VOLUME 7, 2019

A. Menendez-Ortiz et al.: Survey on Reversible Watermarking for Multimedia Content: Robustness Overview
TABLE 9. Self-recovery schemes with approximate reconstruction.
perfect restoration, to construct a framework for reversible
watermarking with watermark and signal robustness; these
works are given in Table 8.
IV. DISCUSSION AND FUTURE WORK
In this manuscript, a review of current state of the art
on reversible watermarking schemes has been presented.
A special focus has been put on the robustness property of
VOLUME 7, 2019
reversible watermarking. Although for many years, reversible
watermarking schemes have been considered inherently frag-
ile, this review demonstrates quite a variety of robustness for
these schemes.
Fragile reversible watermarking schemes were the first
efforts to be proposed, and have been the most popular
reversible schemes for many years, therefore the notion of
the inherent fragility of reversible watermarking. The most
132675
 A. Menendez-Ortiz et al.: Survey on Reversible Watermarking for Multimedia Content: Robustness Overview
commonly used strategy is based on error expansion for both
audio signals and images. Histogram shifting based schemes
are another popular technique for reversible schemes, though
not as widely used as error expansion. Histogram shifting has
been proposed just for images and their results seem to sug-
gest that this type of schemes allow greater payload capacities
than error expansion based schemes. However, it is worth to
mention that schemes based on histogram shifting with the
highest embedding capacities use a multi-level embedding
strategy, whereas single-level embedding produces capacities
much lower than those obtained with error expansion based
schemes. More research has to be done to investigate which
strategy yields a better trade-off between embedding capacity
and transparency, whether multi-bit prediction error expan-
sion or multi-level histogram shifting.
Robustness in reversible watermarking has been explored
mainly for images, with just one scheme for audio signals.
Robust reversible watermarking schemes can be regarded
as semi-fragile or robust, based on the attacks they resist.
Semi-fragile schemes deal only with compression attacks,
such as JPEG and JPEG2000 compression, whereas robust
schemes deal with common signal processing attacks besides
compression. Research is necessary to propose semi-fragile
and robust reversible schemes for other types of multimedia,
such as audio signals and video.
Self-recovery schemes allow the reconstruction of the orig-
inal signals after they have suffered attacks. Though they
cannot be considered reversible watermarking schemes in the
strict sense, since the watermark inserted does not convey
a secret message, they do offer a sort of robustness for the
signal itself, the watermark is inserted using watermarking
techniques and it contains information necessary to counter-
act attacks. Self-recovery schemes with approximate recon-
struction are the schemes where most of the research has
been done, strategies have been proposed for all kinds of
multimedia content, but specially for images. Most of the
strategies for audio signals do not correct attacks, they only
detect modifications and the altered regions of the signal are
identified. More recent efforts have improved the quality of
the reconstructed signals; however, even when these are high
quality signals, they are not a lossless representation of the
host signals. To the best of our knowledge, only five schemes
have been proposed to date that achieve perfect reconstruc-
tion of the host signals, from those, only one scheme has been
designed for audio signals, and the rest deal with images.
However, these schemes resist few attacks, namely content
replacement and cropping. This is an interesting topic, and if
these schemes can resist a wider range of attacks, applications
such as in forensic scenarios could provide not only the iden-
tification of counterfeits, but also what the original contents
were like prior manipulations.
Watermark and signal robustness is a new type of robust-
ness in reversible watermarking, it combines techniques from
fragile reversible watermarking and self-recovery with per-
fect reconstruction. A framework was proposed and it intro-
duced the idea to combine these schemes to work as a
132676
single scheme with both properties, namely, extraction of the
watermark after attacks and restoration of the signals without
loss of data. Two strategies for images and one for audio
signals have been explored, to the best of our knowledge.
The performance of this framework depends on the capac-
ity of the self-recovery scheme to resist attacks; therefore,
the robustness of the whole construction depends on this
capacity.
However, the theoretical capacities for this new kind of
robustness in reversible watermarking opens a new path for
many applications in scenarios with lossy communication
channels. There is much work to be done in this research line.
Robustness, transparency, complexity, and security are just
some of the aspects of these schemes that can be improved
in future works. However, the improvement of this new type
of reversible watermarking schemes can provide interesting
solutions for several applications like in medicine or military
fields.
V. CONCLUSION AND FINAL REMARKS
Robustness is a desired property on digital watermarking
schemes, since in most application scenarios attacks will
occur. Reversible watermarking is an alternative to conven-
tional watermarking when it is necessary to recover the host
signals without loss of data after extraction of the watermark.
However, most of the research in reversible watermarking
has focused, for many years, on reversibility and acceptable
embedding transparency. Therefore, reversible watermarking
has been considered as inherently fragile. Nonetheless, this
manuscript presents a review of reversible schemes with
different types of robustness. A classification based on the
robustness presented by the schemes is proposed and schemes
in each category are compared.
Based on the requirements of the application scenarios,
different watermarking techniques can be used. For instance,
if the main goal is to transmit a secret message but the signal
obtained after extraction tolerates some level of degrada-
tion, conventional watermarking is the best approach. If it is
imperative that the watermarks survive many types of attacks
and a lossless quality of the signal is not vital, some of
the many conventional robust watermarking schemes can be
selected, depending on the set of attacks to be considered in
the application, i.e. signal processing, geometric, time scale,
etc. However if the application does not tolerate distortion of
any kind for the carrier signal, reversible watermarking must
be used. It is worth to mention that prior the introduction of
the reversible watermarking framework with watermark and
signal robustness, restoration of the host signals without loss
of data after watermark extraction was not possible for any
robust reversible scheme. Semi-fragile and robust reversible
watermarking provide robustness for the watermarks, but
can only guarantee the restoration of the host signals for
no attack scenarios. Self-recovery, on the other hand, allows
the reconstruction of a signal after it has suffered some
attacks, however, these schemes cannot convey a secret mes-
sage. Reversible watermarking with watermark and signal
VOLUME 7, 2019
A. Menendez-Ortiz et al.: Survey on Reversible Watermarking for Multimedia Content: Robustness Overview
robustness is the only strategy that allows the restoration of
both when some attacks occur. However, the robustness of the
latter schemes is limited to content replacement and cropping
attacks only.
As it can be seen, the reversibility requirement imposes
a big restriction to watermarking strategies, and to design
techniques for robust watermarks besides reversibility is not a
trivial task. Watermarks need to be inserted in such a way that
they can survive attacks, but at the same time modifications
caused during embedding must be invertible, which in many
cases can be contradictory requirements.
The capacity to resist attacks of all current robust reversible
strategies, namely robust reversible watermarking, self-
recovery or reversible watermarking with watermark and sig-
nal robustness is limited to non-severe common attacks. Work
needs to be done to improve robustness of these schemes, i.e.
to widen the types of attacks that can be resisted and to resist
more severe attacks. The increase on robustness should not
be taken aside of an acceptable trade-off between payload
capacity and transparency, whose boundaries are determined
by the applications but need to be taken into consideration
for all proposals. Finally but not less important is security,
the secrecy of the messages to be conveyed by these schemes
needs to be properly protected.
REFERENCES
[1] J. Seitz, Digital Watermarking for Digital Media. Arlington, VA, USA:
Information Resources Press, 2005.
[2] M. Kutter and F. Hartung, ‘‘Introduction to watermarking techniques,’’
in Information Hiding Techniques for Steganography and Digital Water-
marking, S. Katzenbeisser and F. A. Petitcolas, Eds., 1st ed. Norwood,
MA, USA: Artech House, 2000.
[3] G. Coatrieux, L. Lecornu, B. Sankur, and C. Roux, ‘‘A review of image
watermarking applications in healthcare,’’ in Proc. 28th Annu. Int. Conf.
IEEE Eng. Med. Biol. Soc. (EMBS), Aug./Sep. 2006, pp. 4691–4694.
[4] P. Singh and R. S. Chadha, ‘‘A survey of digital watermarking techniques,
applications and attacks,’’ Int. J. Eng. Innov. Technol., vol. 2, no. 9,
pp. 165–175, Mar. 2013.
[5] C. I. Podilchuk and E. J. Delp, ‘‘Digital watermarking: Algorithms and
applications,’’ IEEE Signal Process. Mag., vol. 18, no. 4, pp. 33–46,
Jul. 2001.
[6] C.-S. Lu, Multimedia Security: Steganography and Digital Watermarking
Techniques for Protection of Intellectual Property. Hershey, PA, USA:
IGI Global, 2004.
[7] N. Cvejic and T. Seppänen, Digital Audio Watermarking Techniques
and Technologies: Applications and Benchmarks. Hershey, PA, USA:
IGI Global, 2008.
[8] N. Agarwal, A. K. Singh, and P. K. Singh, ‘‘Survey of robust and
imperceptible watermarking,’’ Multimedia Tools Appl., vol. 78, no. 7,
pp. 8603–8633, Apr. 2019. doi: 10.1007/s11042-018-7128-5.
[9] L. Singh, A. K. Singh, and P. K. Singh, ‘‘Secure data hiding tech-
niques: A survey,’’ Multimedia Tools Appl., pp. 1–21, Aug. 2018.
doi: 10.1007/s11042-018-6407-5.
[10] M. M. Boone and W. P. J. de Bruijn, ‘‘Improving speech intelligibility
in teleconferencing by using wave field synthesis,’’ in Proc. Audio Eng.
Soc. Conv., Mar. 2003, p. 5800. [Online]. Available: http://www.aes.org/e-
lib/browse.cfm?elib=12559
[11] J. Loomis, R. Golledge, and R. Klatzky, ‘‘Navigation system for the
blind: Auditory display modes and guidance,’’ Presence, vol. 7, no. 2,
pp. 193–203, Apr. 1998. doi: 10.1162/105474698565677.
[12] M. Cohen, O. N. N. Fernando, T. Nagai, and K. Shimizu, ‘‘‘Back-seat
driver’: Spatial sound for vehicular way-finding and situation aware-
ness,’’ in Proc. Jpn.-China Joint Workshop Frontier Comput. Sci. Tech-
nol. (FCST), Washington, DC, USA, 2006, pp. 109–115. doi: 10.1109/
FCST.2006.1.
VOLUME 7, 2019
[13] T. Poitschke, F. Laquai, and G. Rigoll, ‘‘Guiding a driver’s visual attention
using graphical and auditory animations,’’ in Engineering Psychology and
Cognitive Ergonomics (Lecture Notes in Computer Science), vol. 5639,
D. Harris, Ed. Berlin, Germany: Springer, 2009, pp. 424–433.
[14] D. R. Begault, M. R. Anderson, and M. U. Mcclain, ‘‘Spatially modulated
auditory alerts for aviation,’’ in Proc. Int. Conf. Auditory Display, 2006,
vol. 54, no. 4, pp. 276–282. [Online]. Available: http://www.aes.org/e-
lib/browse.cfm?elib=13676
[15] L. J. Hornbeck, ‘‘Digital Light Processing for high-brightness high-
resolution applications,’’ Proc. SPIE, vol. 3013, pp. 27–40, May 1997.
doi: 10.1117/12.273880.
[16] D. Theodoros, A. Hill, and T. Russell, ‘‘Advances int telehealth for the
delivery of voice therapy: The Australian experience,’’ in Proc. 10th Int.
Conf. Adv. Quant. Laryngol., vol. 8, 2013, p. 15.
[17] A. Khan, A. Siddiqa, S. Munib, and S. A. Malik, ‘‘A recent survey of
reversible watermarking techniques,’’ Inf. Sci., vol. 279, pp. 251–272,
Sep. 2014. [Online]. Available: http://www.sciencedirect.com/science/
article/pii/S0020025514004150
[18] N. Thilagavathi, D. Saravanan, S. Kumarakrishnan, S. Punniakodi,
J. Amudhavel, and U. Prabu, ‘‘A survey of reversible watermarking
techniques, application and attacks,’’ in Proc. ACM Int. Conf. Adv. Res.
Comput. Sci. Eng. Technol. (ICARCSET), New York, NY, USA, 2015,
pp. 37-1–37-7.
[19] G. Boato, M. Azzoni, M. Carli, F. Battisti, and K. O. Egiazarian, ‘‘Differ-
ence expansion and prediction for high bit-rate reversible data hiding,’’
J. Electron. Imag., vol. 21, no. 3, pp. 1–10, 2012.
[20] A. M. Alattar, ‘‘Reversible watermark using difference expansion of
quads,’’ in Proc. IEEE Int. Conf. Acoust., Speech, Signal Process., vol. 3,
May 2004, p. iii-377.
[21] A. M. Alattar, ‘‘Reversible watermark using the difference expansion of
a generalized integer transform,’’ IEEE Trans. Image Process., vol. 13,
no. 8, pp. 1147–1156, Aug. 2004.
[22] E. Chrysochos, E. E. Varsaki, V. Fotopoulos, and A. N. Skodras, ‘‘High
capacity reversible data hiding using overlapping difference expansion,’’
in Proc. 10th Workshop Image Anal. Multimedia Interact. Services,
May 2009, pp. 121–124.
[23] T. Naheed, I. Usman, and A. Dar, ‘‘Lossless data hiding using opti-
mized interpolation error expansion,’’ in Proc. Frontiers Inf. Technol.,
Dec. 2011, pp. 281–286.
[24] C. Dragoi and D. Coltuc, ‘‘Improved rhombus interpolation for reversible
watermarking by difference expansion,’’ in Proc. 20th Eur. Signal
Process. Conf. (EUSIPCO), Aug. 2012, pp. 1688–1692.
[25] T. Naheed and I. Usman, ‘‘Intelligent reversible digital watermarking
technique using interpolation errors,’’ in Proc. 11th Int. Conf. Inf. Sci.,
Signal Process. Appl. (ISSPA), Jul. 2012, pp. 1062–1067.
[26] D. Coltuc, ‘‘Improved embedding for prediction-based reversible water-
marking,’’ IEEE Trans. Inf. Forensics Security, vol. 6, no. 3, pp. 873–882,
Sep. 2011.
[27] I. Caciula and D. Coltuc, ‘‘Capacity control of reversible watermarking by
two-thresholds embedding,’’ in Proc. IEEE Int. Workshop Inf. Forensics
Secur. (WIFS), Dec. 2012, pp. 223–227.
[28] D. Coltuc and A. Tudoroiu, ‘‘Multibit versus multilevel embedding in
high capacity difference expansion reversible watermarking,’’ in Proc.
20th Eur. Signal Process. Conf. (EUSIPCO), Aug. 2012, pp. 1791–1795.
[29] I. Caciula and D. Coltuc, ‘‘Capacity control of reversible watermarking by
two-thresholds embedding: Further results,’’ in Proc. Int. Symp. Signals,
Circuits Syst. (ISSCS), Jul. 2013, pp. 1–4.
[30] I.-C. Dragoi and D. Coltuc, ‘‘Local-prediction-based difference expan-
sion reversible watermarking,’’ IEEE Trans. Image Process., vol. 23,
no. 4, pp. 1779–1790, Apr. 2014.
[31] I.-C. Dragoi and D. Coltuc, ‘‘Adaptive pairing reversible watermarking,’’
IEEE Trans. Image Process., vol. 25, no. 5, pp. 2420–2422, May 2016.
[32] M. van der Veen, F. Bruekers, A. van Leest, and S. Cavin, ‘‘High capacity
reversible watermarking for audio,’’ Proc. SPIE, vol. 5020, pp. 1–11,
Jun. 2003.
[33] X. Huang, A. Nishimura, and I. Echizen, ‘‘A reversible acoustic steganog-
raphy for integrity verification,’’ in Digital Watermarking (Lecture Notes
in Computer Science), vol. 6526, H.-J. Kim, Y. Q. Shi, and M. Barni, Eds.
Berlin, Germany: Springer, 2011, pp. 305–316.
[34] A. Nishimura, ‘‘Reversible and robust audio watermarking based on
spread spectrum and amplitude expansion,’’ in Digital-Forensics and
Watermarking, Y.-Q. Shi, H. J. Kim, F. Pérez-González, and C.-N. Yang,
Eds. Cham, Switzerland: Springer, 2015, pp. 215–229.
132677
 A. Menendez-Ortiz et al.: Survey on Reversible Watermarking for Multimedia Content: Robustness Overview
[35] B. Bradley and A. M. Alattar, ‘‘High-capacity invertible data-hiding algo-
rithm for digital audio,’’ Proc. SPIE, vol. 5681, pp. 789–800, Mar. 2005.
doi: 10.1117/12.586042.
[36] J. J. Garcia-Hernandez, ‘‘Exploring reversible digital watermarking in
audio signals using additive interpolation-error expansion,’’ in Proc. 8th
Int. Conf. Intell. Inf. Hiding Multimedia Signal Process. (IIH-MSP),
Jul. 2012, pp. 142–145.
[37] D. Yan and R. Wang, ‘‘Reversible data hiding for audio based on predic-
tion error expansion,’’ in Proc. Int. Conf. Intell. Inf. Hiding Multimedia
Signal Process., Los Alamitos, CA, USA, 2008, pp. 249–252.
[38] A. Nishimura, ‘‘Reversible audio data hiding using linear prediction and
error expansion,’’ in Proc. 7th Int. Conf. Intell. Inf. Hiding Multimedia
Signal Process. (IIH-MSP), Oct. 2011, pp. 318–321.
[39] A. Nishimura, ‘‘Controlling quality and payload in reversible data hiding
based on modified error expansion for segmental audio waveforms,’’ in
Proc. 8th Int. Conf. Intell. Inf. Hiding Multimedia Signal Process. (IIH-
MSP), Jul. 2012, pp. 110–113.
[40] Y. Huo, S. Xiang, S. Liu, X. Luo, and Z. Bai, ‘‘Reversible audio water-
marking algorithm using non-causal prediction,’’ Wuhan Univ. J. Natural
Sci., vol. 18, no. 5, pp. 455–460, 2013. doi: 10.1007/s11859-013-0956-2.
[41] S. Xiang and Z. Li, ‘‘Reversible audio data hiding algorithm using non-
causal prediction of alterable orders,’’ EURASIP J. Audio, Speech, Music
Process., vol. 2017, no. 1, p. 4, Feb. 2017. doi: 10.1186/s13636-017-
0101-9.
[42] A. Bobeica, I. C. Dragoi, I. Caciula, D. Coltuc, F. Albu, and F. Yang,
‘‘Capacity control for prediction error expansion based audio reversible
data hiding,’’ in Proc. 22nd Int. Conf. Syst. Theory, Control Comput.
(ICSTCC), Oct. 2018, pp. 810–815.
[43] Q. Chen, S. Xiang, and X. Luo, ‘‘Reversible watermarking for audio
authentication based on integer DCT and expansion embedding,’’ in
The International Workshop on Digital Forensics and Watermarking
(Lecture Notes in Computer Science), vol. 7809, Y. Shi, H.-J. Kim, and
F. Pérez-González, Eds. Berlin, Germany: Springer, 2013, pp. 395–409.
[44] J. Tian, ‘‘Reversible data embedding using a difference expansion,’’
IEEE Trans. Circuits Syst. Video Technol., vol. 13, no. 8, pp. 890–896,
Aug. 2003.
[45] D. M. Thodi and J. J. Rodriguez, ‘‘Reversible watermarking by
prediction-error expansion,’’ in Proc. 6th IEEE Southwest Symp. Image
Anal. Interpretation, Mar. 2004, pp. 21–25.
[46] J. P. Burg, ‘‘Maximum entropy spectral analysis,’’ Stanford Exploration
Project. Pub., Stanford Univ., Stanford, CA, USA, 1975, no. 6.
[47] Y. Yan, W. Cao, and S. Li, ‘‘High capacity reversible image authentication
based on difference image watermarking,’’ in Proc. IEEE Int. Workshop
Imag. Syst. Techn. (IST), May 2009, pp. 179–182.
[48] V. Sachnev, H. Kim, S. Suresh, and Y. Q. Shi, ‘‘Reversible watermark-
ing algorithm with distortion compensation,’’ EURASIP J. Adv. Signal
Process., vol. 2010, no. 1, 2011, Art. no. 316820. [Online]. Available:
http://asp.eurasipjournals.com/content/2010/1/316820
[49] M. Fujiyoshi, ‘‘A Histogram shifting-based blind reversible data hiding
method with a histogram peak estimator,’’ in Proc. Int. Symp. Commun.
Inf. Technol. (ISCIT), Oct. 2012, pp. 313–318.
[50] S.-Y. Shin, H.-M. Yoo, Y.-H. Ko, H.-S. Kang, and J.-W. Suh, ‘‘Reversible
watermarking without underflow and overflow problems,’’ in Proc.
IEEE 55th Int. Midwest Symp. Circuits Syst. (MWSCAS), Aug. 2012,
pp. 980–983.
[51] Y.-Y. Tsai, D.-S. Tsai, and C.-L. Liu, ‘‘Reversible data hiding
scheme based on neighboring pixel differences,’’ Digit. Signal Process.,
vol. 23, no. 3, pp. 919–927, 2013. [Online]. Available: http://www.
sciencedirect.com/science/article/pii/S1051200412003132
[52] H. T. Wu, J. Huang, Y. Zhang, and J. You, ‘‘Reversible image water-
marking by rhombus prediction and histogram modification,’’ in Proc.
Int. Conf. Audio, Lang. Image Process. (ICALIP), Jul. 2012, pp. 165–169.
[53] G. Coatrieux, W. Pan, N. Cuppens-Boulahia, F. Cuppens, and C. Roux,
‘‘Reversible watermarking based on invariant image classification and
dynamic histogram shifting,’’ IEEE Trans. Inf. Forensics Security, vol. 8,
no. 1, pp. 111–120, Jan. 2013.
[54] L.-C. Huang, L.-Y. Tseng, and M.-S. Hwang, ‘‘A reversible data hid-
ing method by histogram shifting in high quality medical images,’’
J. Syst. Softw., vol. 86, no. 3, pp. 716–727, 2013. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0164121212003172
[55] Z.-H. Wang, C.-F. Lee, and C.-Y. Chang, ‘‘Histogram-shifting-imitated
reversible data hiding,’’ J. Syst. Softw., vol. 86, no. 2, pp. 315–323,
2013. [Online]. Available: http://www.sciencedirect.com/science/article/
pii/S0164121212002415
132678
[56] X.-T. Wang, C.-C. Chang, T.-S. Nguyen, and M.-C. Li, ‘‘Reversible
data hiding for high quality images exploiting interpolation and
direction order mechanism,’’ Digit. Signal Process., vol. 23, no. 2,
pp. 569–577, 2013. [Online]. Available: http://www.sciencedirect.com/
science/article/pii/S1051200412001789
[57] Kamran, A. Khan, and S. A. Malik, ‘‘A high capacity reversible
watermarking approach for authenticating images: Exploiting down-
sampling, histogram processing, and block selection,’’ Inf. Sci., vol. 256,
pp. 162–183, Jan. 2014. [Online]. Available: http://www.sciencedirect.
com/science/article/pii/S0020025513005501
[58] C. W. Honsinger, P. W. Jones, M. Rabbani, and J. C. Stoffel,
‘‘Lossless recovery of an original image containing embedded data,’’
U.S. Patent 6 278 791, Aug. 21, 2001.
[59] C. De Vleeschouwer, J. F. Delaigle, and B. Macq, ‘‘Circular interpre-
tation of bijective transformations in lossless watermarking for media
asset management,’’ IEEE Trans. Multimedia, vol. 5, no. 1, pp. 97–105,
Mar. 2003.
[60] Z. Ni, Y. Q. Shi, N. Ansari, W. Su, Q. Sun, and X. Lin, ‘‘Robust lossless
image data hiding,’’ in Proc. IEEE Int. Conf. Multimedia Expo (ICME),
vol. 3, Jun. 2004, pp. 2199–2202.
[61] Z. Ni, Y. Q. Shi, N. Ansari, W. Su, Q. Sun, and X. Lin, ‘‘Robust loss-
less image data hiding designed for semi-fragile image authentication,’’
IEEE Trans. Circuits Syst. Video Technol., vol. 18, no. 4, pp. 497–509,
Apr. 2008.
[62] D. Zou, Y. Q. Shi, and Z. Ni, ‘‘A semi-fragile lossless digital water-
marking scheme based on integer wavelet transform,’’ in Proc. IEEE 6th
Workshop Multimedia Signal Process., Sep. 2004, pp. 195–198.
[63] D. Zou, Y. Q. Shi, Z. Ni, and W. Su, ‘‘A semi-fragile lossless digi-
tal watermarking scheme based on integer wavelet transform,’’ IEEE
Trans. Circuits Syst. Video Technol., vol. 16, no. 10, pp. 1294–1300,
Oct. 2006.
[64] X. Wu, ‘‘Reversible semi-fragile watermarking based on histogram shift-
ing of integer wavelet coefficients,’’ in Proc. Inaugural IEEE Int. Conf.
Digit. Ecosyst. Technol., Feb. 2007, pp. 501–505.
[65] K.-S. Kim, M.-J. Lee, Y.-H. Suh, and H.-K. Lee, ‘‘Robust lossless data
hiding based on block gravity center for selective authentication,’’ in Proc.
IEEE Int. Conf. Multimedia Expo, Jun./Jul. 2009, pp. 1022–1025.
[66] M. A. Alavianmehr, M. Rezaei, M. S. Helfroush, and A. Tashk, ‘‘A semi-
fragile lossless data hiding scheme based on multi-level histogram shift
in image integer wavelet transform domain,’’ in Proc. 6th Int. Symp.
Telecommun. (IST), Nov. 2012, pp. 976–981.
[67] R. T. Mohammed and B. E. Khoo, ‘‘Robust reversible watermarking
scheme based on wavelet-like transform,’’ in Proc. IEEE Int. Conf. Signal
Image Process. Appl., Oct. 2013, pp. 354–359.
[68] S. Yu, J. Li, and J. Wang, ‘‘A novel robust reversible watermarking
scheme based on IWT,’’ in Cloud Computing and Security, X. Sun,
H.-C. Chao, X. You, and E. Bertino, Eds. Cham, Switzerland: Springer,
2017, pp. 39–48.
[69] E. Chrysochos, V. Fotopoulos, A. Skodras, and M. Xenos, ‘‘Reversible
image watermarking based on histogram modification,’’ in Proc. 11th
Panhellenic Conf. Informat. (PCI), 2007, pp. 93–104.
[70] D. Coltuc and J. M. Chassery, ‘‘Very fast watermarking by reversible
contrast mapping,’’ IEEE Signal Process. Lett., vol. 14, no. 4,
pp. 255–258, Apr. 2007.
[71] G. Coatrieux, J. Montagner, H. Huang, and C. Roux, ‘‘Mixed reversible
and RONI watermarking for medical image reliability protection,’’ in
Proc. 29th Annu. Int. Conf. IEEE Eng. Med. Biol. Soc., Aug. 2007,
pp. 5653–5656.
[72] T.-G. Gao and Q.-L. Gu, ‘‘Reversible watermarking algorithm based
on wavelet lifting scheme,’’ in Proc. Int. Conf. Wavelet Anal. Pattern
Recognit., vol. 4, Nov. 2007, pp. 1771–1775.
[73] M. J. Saberian, M. A. Akhaee, and F. Marvasti, ‘‘An invertible quanti-
zation based watermarking approach,’’ in Proc. IEEE Int. Conf. Acoust.,
Speech Signal Process., Mar./Apr. 2008, pp. 1677–1680.
[74] T.-G. Gao, Q.-L. Gu, and G.-L. Han, ‘‘A novel adaptive reversible water-
mark algorithm based on wavelet lifting scheme,’’ in Proc. Int. Conf. Inf.
Eng. Comput. Sci., 2009, pp. 1–4.
[75] C. C. Chang, P. Y. Lin, and J. S. Yeh, ‘‘Preserving robustness and
removability for digital watermarks using subsampling and difference
correlation,’’ Inf. Sci., vol. 179, no. 13, pp. 2283–2293, Jun. 2009.
[76] C.-Y. Yang, C.-H. Lin, and W.-C. Hu, ‘‘Reversible watermarking by
coefficient adjustment method,’’ in Proc. 6th Int. Conf. Intell. Inf. Hiding
Multimedia Signal Process. (IIH-MSP), 2010, pp. 39–42.
VOLUME 7, 2019
A. Menendez-Ortiz et al.: Survey on Reversible Watermarking for Multimedia Content: Robustness Overview
[77] H.-H. Tsai, H.-C. Tseng, and Y.-S. Lai, ‘‘Robust lossless image
watermarking based on α-trimmed mean algorithm and support
vector machine,’’ J. Syst. Softw., vol. 83, no. 6, pp. 1015–1028, 2010.
[Online]. Available: http://www.sciencedirect.com/science/article/pii/
S0164121209003392
[78] X.-T. Zeng, L.-D. Ping, and X.-Z. Pan, ‘‘A lossless robust data hiding
scheme,’’ Pattern Recognit., vol. 43, no. 4, pp. 1656–1667, Apr. 2010.
[Online]. Available: http://www.sciencedirect.com/science/article/pii/
S0031320309003604
[79] X. Gao, L. An, Y. Yuan, D. Tao, and X. Li, ‘‘Lossless data embedding
using generalized statistical quantity histogram,’’ IEEE Trans. Circuits
Syst. Video Technol., vol. 21, no. 8, pp. 1061–1070, Aug. 2011.
[80] M. A. Hernandez-Morales, ‘‘Esquema robusto de marca de agua digital
reversible en imágenes,’’ M.S. thesis, Nat. Inst. Astrophysics, Optics
Electron., San Andrés Cholula, Mexico, 2012.
[81] L. An, X. Gao, Y. Yuan, D. Tao, C. Deng, and F. Ji, ‘‘Content-adaptive reli-
able robust lossless data embedding,’’ Neurocomputing, vol. 79, pp. 1–11,
Mar. 2012. [Online]. Available: http://www.sciencedirect.com/science/
article/pii/S0925231211005157
[82] L. An, X. Gao, Y. Yuan, and D. Tao, ‘‘Robust lossless data hiding
using clustering and statistical quantity histogram,’’ Neurocomputing,
vol. 77, no. 1, pp. 1–11, Feb. 2012. [Online]. Available: http://www.
sciencedirect.com/science/article/pii/S0925231211003614
[83] L. An, X. Gao, X. Li, D. Tao, C. Deng, and J. Li, ‘‘Robust reversible
watermarking via clustering and enhanced pixel-wise masking,’’ IEEE
Trans. Image Process., vol. 21, no. 8, pp. 3598–3611, Aug. 2012.
[84] Q. Gu and T. Gao, ‘‘A novel reversible robust watermarking algo-
rithm based on chaotic system,’’ Digit. Signal Process., vol. 23,
no. 1, pp. 213–217, 2013. [Online]. Available: http://www.sciencedirect.
com/science/article/pii/S1051200412001674
[85] R. Thabit and B. E. Khoo, ‘‘Robust reversible watermarking scheme
using slantlet transform matrix,’’ J. Syst. Softw., vol. 88, pp. 74–86,
Feb. 2014. [Online]. Available: http://www.sciencedirect.com/science/
article/pii/S0164121213002380
[86] C.-Y. Yang and W.-F. Wang, ‘‘Robust reversible colour image digital
watermarking using offset coefficient technique,’’ in Genetic and Evo-
lutionary Computing, T. T. Zin, J. C.-W. Lin, J.-S. Pan, P. Tin, and
M. Yokota, Eds. Cham, Switzerland: Springer, 2016, pp. 83–93.
[87] J. Fridrich and M. Goljan, ‘‘Images with self-correcting capabilities,’’ in
Proc. Int. Conf. Image Process. (ICIP), vol. 3, 1999, pp. 792–796.
[88] J. Fridrich and M. Goljan, ‘‘Protection of digital images using self-
embedding,’’ in Proc. Symp. Content Secur. Data Hiding Digit. Media.
Newark, NJ, USA, 1999.
[89] B. G. Mobasseri, ‘‘A spatial digital video watermark that survives
MPEG,’’ in Proc. Int. Conf. Inf. Technol., Coding Comput., 2000,
pp. 68–73.
[90] E. Gomez, P. Cano, L. Gomes, E. Batlle, and M. Bonnet, ‘‘Mixed
watermarking-fingerprinting approach for integrity verification of audio
recordings,’’ in Proc. Int. Telecommun. Symp., 2002.
[91] F. Chen, H. He, and H. Wang, ‘‘A fragile watermarking scheme for audio
detection and recovery,’’ in Proc. Congr. Image Signal Process., vol. 5,
2008, pp. 135–138.
[92] C. Qin, C.-C. Chang, and P.-Y. Chen, ‘‘Self-embedding fragile water-
marking with restoration capability based on adaptive bit alloca-
tion mechanism,’’ Signal Process., vol. 92, no. 4, pp. 1137–1150,
2012. [Online]. Available: http://www.sciencedirect.com/science/article/
pii/S0165168411003987
[93] P. Korus and A. Dziech, ‘‘Reconfigurable self-embedding with high
quality restoration under extensive tampering,’’ in Proc. IEEE Int. Conf.
Image Process. (ICIP), Sep./Oct. 2012, pp. 2193–2196.
[94] S. Dadkhah, A. A. Manaf, Y. Hori, A. E. Hassanien, and S. Sadeghi,
‘‘An effective SVD-based image tampering detection and self-recovery
using active watermarking,’’ Signal Process., Image Commun.,
vol. 29, no. 10, pp. 1197–1210, 2014. [Online]. Available: http://www.
sciencedirect.com/science/article/pii/S0923596514001313
[95] X. Zhang and S. Wang, ‘‘Fragile watermarking with error-free restoration
capability,’’ IEEE Trans. Multimedia, vol. 10, no. 8, pp. 1490–1499,
Dec. 2008.
[96] X. Zhang, S. Wang, Z. Qian, and G. Feng, ‘‘Reference sharing mechanism
for watermark self-embedding,’’ IEEE Trans. Image Process., vol. 20,
no. 2, pp. 485–495, Feb. 2011.
[97] S. Bravo-Solorio, C.-T. Li, and A. Nandi, ‘‘Watermarking method with
exact self-propagating restoration capabilities,’’ in Proc. IEEE Int. Work-
shop Inf. Forensics Secur. (WIFS), Dec. 2012, pp. 217–222.
VOLUME 7, 2019
[98] S. Bravo-Solorio, F. Calderon, C.-T. Li, and A. K. Nandi, ‘‘Fast frag-
ile watermark embedding and iterative mechanism with high self-
restoration performance,’’ Digit. Signal Process., vol. 73, pp. 83–92,
Feb. 2018. [Online]. Available: http://www.sciencedirect.com/science/
article/pii/S1051200417302658
[99] A. Menéndez-Ortiz, C. Feregrino-Uribe, J. J. Garcia-Hernandez, ‘‘Self-
recovery scheme for audio restoration using auditory masking,’’ PLoS
ONE, vol. 13, no. 9, 2018, Art. no. e0204442. doi: 10.1371/journal.
pone.0204442.
[100] T. Kalker and F. M. J. Willems, ‘‘Capacity bounds and constructions
for reversible data-hiding,’’ in Security and Watermarking of Multimedia
Contents V, vol. 5020. Bellingham, WA, USA: SPIE, 2003, pp. 604–611.
doi: 10.1117/12.473164.
[101] F. Willems and T. Kalker, ‘‘Coding theorems for reversible embedding,’’
DIMACS Ser. Discrete Math. Theor. Comput. Sci., vol. 66, pp. 61–78,
Aug. 2004.
[102] S. Kotagiri and J. N. Laneman, ‘‘Reversible information embedding
in multi-user channels,’’ in Proc. Allerton Conf. Commun., Control,
Comput., 2005, pp. 1–10.
[103] S. Kotagiri and J. N. Laneman, ‘‘Information embedding in degraded
broadcast channels,’’ in Proc. IEEE Int. Symp. Inf. Theory, Jul. 2006,
pp. 494–498.
[104] Y. Steinberg, ‘‘Reversible information embedding with compressed host
at the decoder,’’ in Proc. IEEE Int. Symp. Inf. Theory, Jul. 2006,
pp. 188–191.
[105] Y. Steinberg, ‘‘Coding for channels with rate-limited side information at
the decoder, with applications,’’ IEEE Trans. Inf. Theory, vol. 54, no. 9,
pp. 4283–4295, Sep. 2008.
[106] Y. Steinberg, ‘‘Simultaneous transmission of data and state with common
knowledge,’’ in Proc. IEEE Int. Symp. Inf. Theory, Jul. 2008, pp. 935–939.
[107] Y. Steinberg, ‘‘Coding and common reconstruction,’’ IEEE Trans. Inf.
Theory, vol. 55, no. 11, pp. 4995–5010, Nov. 2009.
[108] O. Sumszyk and Y. Steinberg, ‘‘Information embedding with reversible
stegotext,’’ in Proc. IEEE Int. Symp. Inf. Theory, Jun./Jul. 2009,
pp. 2728–2732.
[109] W. Zhang, B. Chen, and N. Yu, ‘‘Capacity-approaching codes for
reversible data hiding,’’ in Information Hiding (Lecture Notes in Com-
puter Science), vol. 6958, T. Filler, T. Pevný, S. Craver, and A. Ker, Eds.
Berlin, Germany: Springer, 2011, pp. 255–269.
[110] W. Zhang, B. Chen, and N. Yu, ‘‘Improving various reversible data
hiding schemes via optimal codes for binary covers,’’ IEEE Trans. Image
Process., vol. 21, no. 6, pp. 2991–3003, Jun. 2012.
[111] W. Zhang, X. Hu, X. Li, and N. Yu, ‘‘Recursive histogram modifica-
tion: Establishing equivalency between reversible data hiding and loss-
less data compression,’’ IEEE Trans. Image Process., vol. 22, no. 7,
pp. 2775–2785, Jul. 2013.
[112] A. Menéndez-Ortiz, C. Feregrino-Uribe, and J. J. García-Hernández,
‘‘Reversible image watermarking scheme with perfect watermark and
host restoration after a content replacement attack,’’ in Proc. Int. Conf.
Secur. Manage. (SAM), vol. 13, Jul. 2014, pp. 385–391.
[113] D. W. Robledo-Cruz, ‘‘Esquema de mercado de agua digital reversible
robusto a reemplazo de contenido para imágenes digítales,’’ M.S. thesis,
CINVESTAV, Mexico City, Mexico, Oct. 2015.
[114] A. Menéndez-Ortiz, ‘‘Reversible watermarking scheme with wa-
termark and signal robustness for audio,’’ Ph.D. dissertation, INAOE,
San Andrés Cholula, Mexico, Jun. 2017.
[115] A. Menéndez-Ortiz, C. Feregrino-Uribe, and J. J. Garcia-Hernandez,
‘‘Audio reversible watermarking scheme in the intDCT domain with
modified prediction error expansion,’’ in Proc. 5th ACM Workshop Inf.
Hiding Multimedia Secur. (IH MMSec), New York, NY, USA, 2017,
pp. 105–110. doi: 10.1145/3082031.3083246.
[116] B. G. Mobasseri and A. T. Evans, ‘‘Content-dependent video authentica-
tion by self-watermarking in color space,’’ Proc. SPIE, vol. 4314, no. 35,
pp. 35–44, Aug. 2001. doi: 10.1117/12.435437.
[117] H.-C. Wu and C.-C. Chang, ‘‘Detection and restoration of tampered
JPEG compressed images,’’ J. Syst. Softw., vol. 64, no. 2, pp. 151–161,
2002. [Online]. Available: http://www.sciencedirect.com/science/article/
pii/S016412120200033X
[118] M. U. Celik, G. Sharma, A. M. Tekalp, and E. S. Saber, ‘‘Video
authentication with self-recovery,’’ Proc. SPIE, vol. 4675, pp. 531–541,
Apr. 2002. doi: 10.1117/12.465311.
[119] M. Steinebach and J. Dittmann, ‘‘Watermarking-based digital audio data
authentication,’’ EURASIP J. Adv. Signal Process., vol. 2003, Dec. 2003,
Art. no. 252490. doi: 10.1155/S1110865703304081.
132679
 A. Menendez-Ortiz et al.: Survey on Reversible Watermarking for Multimedia Content: Robustness Overview
[120] R. Caldelli, F. Bartolini, V. Cappellini, A. Piva, and M. Barni, ‘‘A new
self-recovery technique for image authentication,’’ in Visual Content
Processing and Representation (Lecture Notes in Computer Science),
vol. 2849, N. García, L. Salgado, and J. M. Martínez, Eds. Berlin,
Germany: Springer, 2003, pp. 164–171.
[121] P. L. Lin, C.-K. Hsieh, and P.-W. Huang, ‘‘A hierarchical digital water-
marking method for image tamper detection and recovery,’’ Pattern
Recognit., vol. 38, no. 12, pp. 2519–2529, 2005. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0031320305000890
[122] X. Zhu, A. T. S. Ho, and P. Marziliano, ‘‘A new semi-fragile image
watermarking with robust tampering restoration using irregular sam-
pling,’’ Signal Process., Image Commun., vol. 22, no. 5, pp. 515–528,
2007. [Online]. Available: http://www.sciencedirect.com/science/article/
pii/S0923596507000434
[123] X. Zhao, A. T. S. Ho, H. Treharne, V. Pankajakshan, C. Culnane, and
W. Jiang, ‘‘A novel semi-fragile image watermarking, authentication
and self-restoration technique using the slant transform,’’ in Proc. 3rd
Int. Conf. Intell. Inf. Hiding Multimedia Signal Process., vol. 1, 2007,
pp. 283–286.
[124] M.-S. Wang and W.-C. Chen, ‘‘A majority-voting based watermarking
scheme for color image tamper detection and recovery,’’ Comput. Stan-
dards Interfaces, vol. 29, no. 5, pp. 561–570, 2007. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0920548906001346
[125] K. L. Hung and C.-C. Chang, ‘‘Recoverable tamper proofing technique
for image authentication using irregular sampling coding,’’ in Autonomic
and Trusted Computing (Lecture Notes in Computer Science), vol. 4610,
B. Xiao, L. T. Yang, J. Ma, C. Muller-Schloer, and Y. Hua, Eds. Berlin,
Germany: Springer, 2007, pp. 333–343.
[126] Y. M. Y. Hasan and A. M. Hassan, ‘‘Tamper detection with self-correction
hybrid spatial-DCT domains image authentication technique,’’ in Proc.
IEEE Int. Symp. Signal Process. Inf. Technol., Dec. 2007, pp. 369–374.
[127] S.-S. Wang and S.-L. Tsai, ‘‘Automatic image authentication and
recovery using fractal code embedding and image inpainting,’’ Pat-
tern Recognit., vol. 41, no. 2, pp. 701–712, 2008. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0031320307002518
[128] X. Jiang and Q. Liu, ‘‘Semi-fragile watermarking algorithm for image
tampers localization and recovery,’’ J. Electron., China, vol. 25, no. 3,
pp. 343–351, 2008. doi: 10.1007/s11767-006-6209-1.
[129] H. He, J. Zhang, and F. Chen, ‘‘A self-recovery fragile watermark-
ing scheme for image authentication with superior localization,’’ Sci.
China F, Inf. Sci., vol. 51, no. 10, pp. 1487–1507, 2008. doi: 10.1007/
s11432-008-0094-1.
[130] A. Karantonis and J. Ellinas, ‘‘Self embedding watermarking for prevent-
ing image content from tampering,’’ Technol. Educ. Inst. Piraeus, Egaleo,
Greece, Tech. Rep., 2009.
[131] H.-J. He, J.-S. Zhang, and H.-M. Tai, ‘‘Self-recovery fragile watermarking
using block-neighborhood tampering characterization,’’ in Information
Hiding (Lecture Notes in Computer Science), vol. 5806, S. Katzenbeisser
and A.-R. Sadeghi, Eds. Berlin, Germany: Springer, 2009, pp. 132–145.
[132] A. M. Hassan, A. Al-Hamadi, Y. M. Y. Hasan, M. A. A. Wahab, and
B. Michaelis, ‘‘Secure block-based video authentication with localization
and self-recovery,’’ World Acad. Sci., Eng. Technol., vol. 2009, no. 33,
pp. 69–74, Sep. 2009.
[133] A. Cheddad, J. Condell, K. Curran, and P. M. Kevitt, ‘‘A secure
and improved self-embedding algorithm to combat digital docu-
ment forgery,’’ Signal Process., vol. 89, no. 12, pp. 2324–2332,
2009. [Online]. Available: http://www.sciencedirect.com/science/article/
pii/S0165168409000425
[134] Z. Qian and G. Feng, ‘‘Inpainting assisted self recovery with decreased
embedding data,’’ IEEE Signal Process. Lett., vol. 17, no. 11,
pp. 929–932, Nov. 2010.
[135] Z. Qian and T. Qiao, ‘‘Image self-embedding with large-area restoration
capability,’’ in Proc. Int. Conf. Multimedia Inf. Netw. Secur. (MINES),
2010, pp. 649–652.
[136] J. A. Mendoza-Noriega, B. M. Kurkoski, M. Nakano-Miyatake, and
H. Perez-Meana, ‘‘Halftoning-based self-embedding watermarking for
image authentication and recovery,’’ in Proc. IEEE Int. Midwest Symp.
Circuits Syst. (MWSCAS), Aug. 2010, pp. 612–615.
[137] P. Korus, W. Szmuc, and A. Dziech, ‘‘A scheme for censorship of sensitive
image content with high-quality reconstruction ability,’’ in Proc. IEEE
Int. Conf. Multimedia Expo (ICME), Jul. 2010, pp. 1073–1078.
[138] M. Iwata, T. Hori, A. Shiozaki, and A. Ogihara, ‘‘Digital watermarking
method for tamper detection and recovery of JPEG images,’’ in Proc. Int.
Symp. Inf. Theory Appl. (ISITA), 2010, pp. 309–314.
132680
[139] A. M. Hassan, A. Al-Hamadi, B. Michaelis, Y. M. Y. Hasan, and
M. A. A. Wahab, ‘‘Secure self-recovery image authentication using
randomly-sized blocks,’’ in Proc. 20th Int. Conf. Pattern Recognit.
(ICPR), 2010, pp. 1445–1448.
[140] A. M. Hassan, A. Al-Hamadi, Y. M. Y. Hasan, M. A. A. Wahab,
A. Panning, and B. Michaelis, ‘‘Variable block-size image authentication
with localization and self-recovery,’’ in Proc. 17th IEEE Int. Conf. Image
Process. (ICIP), Sep. 2010, pp. 3665–3668.
[141] C. Cruz, R. Reyes, M. Nakano, and H. Pérez, ‘‘Semi-fragile watermarking
based content image authentication scheme,’’ Revista Facultad Ingeniería
Univ. Antioquia, vol. 56, pp. 160–169, Dec. 2010.
[142] X. Zhang, Z. Qian, Y. Ren, and G. Feng, ‘‘Watermarking with flexi-
ble self-recovery quality based on compressive sensing and composi-
tive reconstruction,’’ IEEE Trans. Inf. Forensics Security, vol. 6, no. 4,
pp. 1223–1232, Dec. 2011.
[143] Y. Shi, M. Qi, Y. Lu, J. Kong, and D. Li, ‘‘Object based self-embedding
watermarking for video authentication,’’ in Proc. Int. Conf. Transp.,
Mech., Elect. Eng. (TMEE), 2011, pp. 519–522.
[144] J. A. Mendoza-Noriega, M. Kurkoski, M. N. Miyatake, and H. P. Meana,
‘‘Image authentication and recovery using BCH error-correcting codes,’’
Int. J. Comput., vol. 5, no. 1, pp. 26–33, 2011.
[145] C. Li, Y. Wang, B. Ma, and Z. Zhang, ‘‘A novel self-recovery fragile
watermarking scheme based on dual-redundant-ring structure,’’ Comput.
Electr. Eng., vol. 37, no. 6, pp. 927–940, 2011. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0045790611001327
[146] P. Korus, L. Janowski, and P. Romaniak, ‘‘Automatic quality control of
digital image content reconstruction schemes,’’ in Proc. IEEE Int. Conf.
Multimedia Expo (ICME), Jul. 2011, pp. 1–6.
[147] H. He, F. Chen, H.-M. Tai, T. Kalker, and J. Zhang, ‘‘Performance anal-
ysis of a block-neighborhood-based self-recovery fragile watermarking
scheme,’’ IEEE Trans. Inf. Forensics Security, vol. 7, no. 1, pp. 185–196,
Feb. 2012.
[148] S. Bravo-Solorio, C.-T. Li, and A. K. Nandi, ‘‘Watermarking with lowem-
bedding distortion and self-propagating restoration capabilities,’’ in Proc.
IEEE Int. Conf. Image Process. (ICIP), Sep./Oct. 2012, pp. 2197–2200.
[149] T. Ahsan, T. Mohammad, M. S. Alam, and U.-P. Chong, ‘‘Digital water-
marking based image authentication and restoration by quantization
of integer wavelet transform coefficients,’’ in Proc. Int. Conf. Inform.,
Electron. Vis. (ICIEV), 2012, pp. 1163–1167.
[150] Y. Shi, M. Qi, Y. Yi, M. Zhang, and J. Kong, ‘‘Object based dual
watermarking for video authentication,’’ Optik-Int. J. Light Electron
Opt., vol. 124, no. 19, pp. 3827–3834, 2013. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0030402613001113
[151] P. Korus and A. Dziech, ‘‘Efficient method for content reconstruction
with self-embedding,’’ IEEE Trans. Image Process., vol. 22, no. 3,
pp. 1134–1147, Mar. 2013.
[152] A. A. Ansari, M. Pant, and C. Ahn, ‘‘SVD based fragile watermark-
ing scheme for tamper localization and self-recovery,’’ Int. J. Mach.
Learn. Cybern., vol. 7, no. 6, pp. 1225–1239, Dec. 2016. doi: 10.1007/
s13042-015-0455-1.
ALEJANDRA MENENDEZ-ORTIZ received the
B.Sc. degree in computer sciences from Uni-
versidad del Mar, in 2010, and the M.Sc.
and Ph.D. degrees in computer sciences from
Instituto Nacional de Astrofísica, Óptica y Elec-
trónica (INAOE), in 2012 and 2017, respec-
tively, where she developed robust watermarking
schemes for audio signals and audio watermarking
schemes with restoration capabilities and water-
mark robustness. She is currently a Postdoctoral
Fellow with the Department of Computer Science, Center for Research
in Mathematics (CIMAT-Mexico). Her research interests focus on water-
marking schemes for audio, reversible watermarking, and robust reversible
watermarking for both images and audio.
VOLUME 7, 2019
A. Menendez-Ortiz et al.: Survey on Reversible Watermarking for Multimedia Content: Robustness Overview

CLAUDIA FEREGRINO-URIBE received the JOSE JUAN GARCIA-HERNANDEZ received the

B.Sc. degree in computer systems engineering
from the Instituto Tecnológico de Querétaro, Mex-
ico, the M.Sc. degree in electrical engineering
with telecommunications option from the Cen-
tro de Investigación y de Estudios Avanzados
del Instituto Politécnico Nacional (CINVESTAV-
IPN), Guadalajara, Mexico, in 1997, and the Ph.D.
degree in electronic engineering in digital systems
from Loughborough University, U.K., in 2001.
She is currently a Researcher with the Computer Science Department,
Instituto Nacional de Astrofísica, Óptica y Electrónica (INAOE), Mexico.
Her research areas are datacompression, cryptography and steganography
(Watermarking), digital systems design, FPGA applications, and medical
imaging applications.
B.Sc. degree in communications and electronic
engineering, the M.Sc. degree in microelectron-
ics, and the Ph.D. degree in communications
and electronics from Escuela Superior de Inge-
niería, Mecánica y Eléctrica del Instituto Politec-
nico Nacional (ESIME-IPN), Mexico, in 2002,
2004, and 2008, respectively. From 2008 to 2010,
he was a Postdoctoral Fellow with the Insti-
tuto Nacional de Astrofísica, Óptica y Electrónica
(INAOE), Mexico. He is currently an Associate Professor with the Centro
de Investigación y de Estudios Avanzados del Instituto Politécnico Nacional
(CINVESTAV-IPN), Tamaulipas, Mexico. His research interests include
multimedia signal processing for data hiding applications and FPGA-based
reconfigurable computing for DSP.

ROGELIO HASIMOTO-BELTRAN received the

Ph.D. degree in computer and electrical engi-
neering from the University of Delaware, USA,
in 2001. He spent two years with Akamai Tech-
nologies (a leader enterprise in multimedia content
delivery), as a Senior Software Engineer. He was
a Visiting Associate Professor with the Univer-
sity of Illinois at Chicago, from 2009 to 2010.
He has been a Senior Scientific Researcher with
the Department of Computer Science, Center for
Research in Mathematics (CIMAT-Mexico), since 2003. He has published
more than 40 technical articles in refereed conferences and journals in the
area of image processing, computer vision, and multimedia networks. His
current research interests include robust multimedia communication, error
concealment, face detection and recognition, and chaotic encryption. He is a
member of the IEEE Association.

VOLUME 7, 2019 132681