CHAPTER 1 INTRODUCTION 1.

Introduction to Digital Image Processing:
   Vision allows humans to perceive and understand the world surrounding us. Computer vision aims to duplicate the effect of human vision by electronically perceiving and understanding an image. Giving computers the ability to see is not an easy task - we live in a three dimensional (3D) world, and when computers try to analyze objects in 3D space, available visual sensors (e.g., TV cameras) usually give two dimensional (2D) images, and this projection to a lower number of dimensions incurs an enormous loss of information.    In order to simplify the task of computer vision understanding, two levels are usually distinguished; low-level image processing and high level image understanding. Usually very little knowledge about the content of images High level processing is based on knowledge, goals, and plans of how to achieve those goals. Artificial intelligence (AI) methods are used in many cases. High-level computer vision tries to imitate human recognition and the ability to make decisions according to the information contained in the image.   This course deals almost exclusively with low-level image processing, high level in which is a continuation of this course. Age processing is discussed in the course Image Analysis and Understanding, which is a continuation of this course.

1.1 History:
Many of the techniques of digital image processing, or digital picture processing as it was often called, were developed in the 1960s at the Jet Propulsion Laboratory, MIT, Bell Labs, University of Maryland, and few other places, with application to satellite imagery, wire photo standards conversion, medical imaging, videophone, character recognition, and photo enhancement. But the cost of processing was fairly high with the computing equipment of that era. In the 1970s, digital image processing proliferated, when cheaper computers Creating a film or electronic image of any picture or paper form. It is accomplished by scanning or photographing an object and turning it into a matrix of dots (bitmap), the meaning of which is unknown to the computer, only to the human

viewer. Scanned images of text may be encoded into computer data (ASCII or EBCDIC) with page recognition software (OCR).

1.2 Basic Concepts:
 

A signal is a function depending on some variable with physical meaning. Signals can be
o o o o

One-dimensional (e.g., dependent on time), Two-dimensional (e.g., images dependent on two co-ordinates in a plane), Three-dimensional (e.g., describing an object in space), Or higher dimensional.

1.3 Pattern recognition
Pattern recognition is a field within the area of machine learning. Alternatively, it can be defined as "the act of taking in raw data and taking an action based on the category of the data" . As such, it is a collection of methods for supervised learning. Pattern recognition aims to classify data (patterns) based on either a priori knowledge or on statistical information extracted from the patterns. The patterns to be classified are usually groups of measurements or observations, defining points in an appropriate multidimensional space. Are to represent, for example, color images consisting of three component colors.

1.4 Image functions:
 

The image can be modeled by a continuous function of two or three variables; Arguments are co-ordinates x, y in a plane, while if images change in time a third variable t might be added.

 

The image function values correspond to the brightness at image points. The function value can express other physical quantities as well (temperature, pressure distribution, distance from the observer, etc.).

The brightness integrates different optical quantities - using brightness as a basic quantity allows us to avoid the description of the very complicated process of image formation.

The image on the human eye retina or on a TV camera sensor is intrinsically 2D. We shall call such a 2D image bearing information about brightness points an intensity image.

  

The real world, which surrounds us, is intrinsically 3D. The 2D intensity image is the result of a perspective projection of the 3D scene. When 3D objects are mapped into the camera plane by perspective projection a lot of information disappears as such a transformation is not one-to-one.

 

Recognizing or reconstructing objects in a 3D scene from one image is an ill-posed problem. Recovering information lost by perspective projection is only one, mainly geometric, problem of computer vision.

The second problem is how to understand image brightness. The only information available in an intensity image is brightness of the appropriate pixel, which is dependent on a number of independent factors such as
o

Object surface reflectance properties (given by the surface material, microstructure and marking),

o O

Illumination properties, And object surface orientation with respect to a viewer and light source.

CHAPTER 2 2. DIGITAL IMAGE FORENSICS
Today's technology allows digital media to be altered and manipulated in ways that were impossible twenty years ago. We are feeling the impact of this technology in nearly every corner of our lives, from the courts to the media, politics, business, and science. As this technology continues to evolve it will become increasingly more important for the science of digital forensics to keep pace. This presentation will describe state of the art techniques in digital image forensics. Digital watermarking has been proposed as a means by which an image can be authenticated. This approach works by inserting at the time of recording an imperceptible digital code (a watermark) into the image. With the assumption that tampering will alter a watermark, an image can be authenticated by verifying that the extracted watermark is the same as that which was inserted. The major drawback of this approach is that a watermark must be inserted at precisely the time of recording, which limits this approach to specially equipped digital cameras. In contrast, recent advances in digital forensics operate in the absence of any watermark or specialized hardware. With the assumption that tampering disturbs certain underlying statistical properties of an image, these forensic techniques can detect specific forms of tampering. Air-brushing or re-touching can be detected by measuring deviations of the underlying color filter array correlations. Specifically, virtually all digital cameras record only a subset of all the pixels needed for a full-resolution color image. Instead, only a subset of the pixels is recorded by a color filter array (CFA) placed atop the digital sensor. The most frequently used CFA, the Bayer array, employs three color filters: red, green, and blue. Since only a single color sample is recorded at each pixel location, the other two color samples must be estimated from the neighboring samples in order to obtain a three-channel color image. The estimation of the missing color samples is referred to as CFA interpolation or demosaicking. In its simplest form, the missing pixels are filled in by spatially averaging the recorded values. Since the CFA is arranged in a periodic pattern, a periodic set of pixels will be precisely correlated to their neighbors according to the CFA interpolation algorithm. When an image is re-touched, it is likely that these correlations will be destroyed. As such, the presence or lack of these correlations can be used to authenticate an image, or expose it as a forgery. A digital composite of two people can be detected by measuring differences in the direction to the illuminating light sources from their faces and body. By making some initial simplifying assumptions about the light and the surface being illuminated, we can mathematically express how much light a surface should receive as a function of its position relative to the light. A surface that is

Any inconsistencies in lighting can then be used as evidence of tampering. Instead. a region growing algorithm combines any significant number of neighboring blocks that are consistent with the cloning of an image region. after using some algorithm. This form of tampering can be detected by first partitioning an image into small blocks. It is prominently used for tracing copyright infringements and for banknote authentication.1. whereas in digital watermarking. it is of no use. Since it is statistically unlikely to find identical and spatially coherent regions in an image. The needed properties of a digital watermark depend on the use case in which it is applied. video. standard techniques can be used to determine the direction to the light source for any object or person in an image. texts or 3D models. Once expressed in this form. 2. but does not need to contain a relation to the carrier signal. With identical and highly similar blocks neighboring each other in the re-ordered sequence. Unlike metadata that is added to the carrier signal. The blocks are then re-ordered so that they are placed a distance to each other that is proportional to the differences in their pixel colors. the hidden information should. Traditional Watermarks may be applied to visible media (like images or video). if integrity has to be ensured. will be brighter than a surface that is turned away from the light. a digital watermark has to be rather robust against modifications that can be applied to the carrier signal.directly facing the light. their presence can then be used as evidence of tampering. Digital watermarks may be used to verify the authenticity or integrity of the carrier signal or to show the identity of its owners. It is typically used to identify ownership of the copyright of such signal. . digital watermarks are only perceptible under certain conditions. pictures. A signal may carry several different watermarks at the same time. Like traditional watermarks. and imperceptible anytime else. the signal may be audio. "Watermarking" is the process of hiding digital information in a carrier signal. for example. If a digital watermark distorts the carrier signal in a way that it becomes perceivable. Duplication or cloning is a simple and powerful form of manipulation used to remove objects or people from an image. a fragile watermark would be applied. i. DIGITAL WATERMARKING A digital watermark is a kind of marker covertly embedded in a noise-tolerant signal such as audio or image data. a digital watermark does not change the size of the carrier signal.e. For marking media files with copyright information.

digital watermarking is a passive protection tool. A watermark is embedded into a digital signal at each point of distribution.1 Principle of digital watermarks A watermark on a bank note has a different transparency than the rest of the note when a light is shined on it. However. . Since a digital copy of data is the same as the original. to authenticate the content.Both steganography and digital watermarking employ steganographic techniques to embed data covertly in noisy signals. can be used later to identify the owner of the work. quality. The media of focus in this paper is the still image. but does not degrade it nor controls access to the data. it should be robust. The effectiveness of a watermark is improved when the technique exploits known properties of the human visual system. ownership. and common signal processing alterations used to make the data more efficient to store and transmit. To be effective in its purpose. and makes detection. In particular. the owner can still be identified. If a copy of the work is found later.2. These are known as perceptually based watermarking techniques. known as a watermark. and source. .1. This technique reportedly has been used to detect the source of illegally copied movies. Within this category. and therefore removal. the class of image-adaptive watermarks proves most effective. by pirates less possible. This is so that afterwards. Robustness requires that it be able to survive any alterations or distortions that the watermarked content may undergo. depending on in which domain the watermark is constructed: the spatial domain (producing spatial watermarks) and the frequency domain (producing spectral watermarks). Digital watermarking is the process of inserting a digital signal or pattern (indicative of the owner of the content) into digital content. falling into 2 main categories. and to trace illegal copies of the work. Transparency requires a watermark to be imperceptible so that it does not affect the quality of the content. including intentional attacks to remove the watermark. this method is useless in the digital world. and transparent. The signal. One application of digital watermarking is source tracking. Watermarks of varying degrees of obtrusiveness are added to presentation media as a guarantee of authenticity. digital watermarking tries to control the robustness as top priority. It just marks data. then the watermark may be retrieved from the copy and the source of the distribution is known. But whereas steganography aims for imperceptibility to human senses. There are a variety of image watermarking techniques. a watermark should adhere to a few requirements.

musicians and artists. photographers.3 PURPOSES OF DIGITAL WATERMARKS Watermarks are a way of dealing with the problems mentioned above by providing a number of services:  They aim to mark digital data permanently and unalterably.1. whether their music is released over the Internet or on DVD-Audio". However. RIAA reports that "record labels see watermarking as a crucial piece of the copy protection system. including writers. image and video databases. Embedding a watermark should not result in a significant increase or reduction in the original data. 2. Copyright owners can incorporate . image's origin and content can be verified. these same opportunities provide ease of access to these works. It is easy to duplicate audio and visual files. Digital watermarking is the content protection method for the multimedia era. surveillance imaging. digital libraries.2 IMPORTANCE OF DIGITAL WATERMARKS The Internet has provided worldwide publishing opportunities to creators of various works.1. and that digital watermarking is needed to indicate who the culprit is. 2. so that the source as well as the intended recipient of the digital work is known. They are of the opinion that any encryption system can be broken.Currently there are various techniques for embedding digital watermarks. and is therefore probable that duplication on the Internet occurs without the rightful owners' permission. An example of an area where copyright protection needs to be enforced is in the on-line music industry. and satellite-delivered video. video-ondemand systems. Furthermore. When digital cameras are used to snapshot an event. Another scenario in which the enforcement of copyright is needed is in newsgathering. The Recording Industry Association of America (RIAA) says that the value of illegal copies of music that are distributed over the Internet could reach $2 billion a year. This suggests that there are many applications that could require image watermarking. Digital watermarking is being recognized as a way for improving this situation. medical imaging. digital cameras. they all digitally write desired information directly onto images or audio data in such a manner that the images or audio data are not damaged. which has resulted in pirating. This is so that later. Digital watermarks are added to images or audio data in such a way that they are invisible or inaudible Ñ unidentifiable by human eye or ear. Basically. sooner or later. they can be embedded in content with a variety of file formats. including Internet imaging. the images must be watermarked as they are captured.

A/D conversion Resampling.5 DIGITAL WATERMARKING APPLICATIONS Digital watermarking is rapid evolving field.  With a tracking service. translation. That is. Requantization.1. scaling and cropping. because each purchaser of the data has a unique watermark embedded in his/her copy. owners are able to find illegal copies of their work on the Internet.  Some more recent techniques are able to correct the alteration as well. any unauthorized copies that s/he has distributed can be traced back to him/her. watermarks are used in the protection of ownership. Addition of a constant offset to the pixel values Local exchange of pixels other intentional attacks: Printing and Rescanning Watermarking of watermarked image (rewatermarking) 2.  By indicating the owner of the work. Geometric Distortions: include such operations as rotation. D/A conversion. Recompression Linear filtering such as high pass and low pass filtering. this section identifies digital watermarking applications and provides an overview of digital watermarking capabilities and useful benefits to customers. In addition.  Watermarks can be used to identify any changes that have been made to the watermarked data.4 ATTACKS ON WATERMARKS            Lossy Compression: Many compression schemes like JPEG and MPEG can potentially degrade the data’s quality through irretrievable loss of data. The presence of a watermark in a work suspected of having been copied can prove that it has been copied.identifying information into their work. 2. The various applications are:    Authentication Broadcast Monitoring Copy Prevention .1. they demonstrate the quality and assure the authenticity of the work. Common Signal Processing Operations: They include the followings.

Digital signatures can also provide nonrepudiation.t. while also claiming their private key remains secret. some non-repudiation schemes offer a time stamp for the digital signature. 2. a broader term that refers to any electronic data that carries the intent of a signature. and must be implemented properly to be effective. copyright protection. MediaSec: Provide software for various media types.[1] but not all electronic signatures use digital signatures. the signature is valid. Digital signatures are commonly used for software distribution.[2][3] In some countries. and in other cases where it is important to detect forgery or tampering. financial transactions. so that even if the private key is exposed. Digital signatures employ a type of asymmetric cryptography. Signum: Allow digital fingerprints to be embedded into grahics. including the United States.[4] and members of the European Union. Stegnosign: For creating. Digital signatures are often used to implement electronic signatures.1. For messages sent through a nonsecure channel. Digital signatures are equivalent to traditional handwritten signatures in many respects. . but properly implemented digital signatures are more difficult to forge than the handwritten type. electronic signatures have legal significance.c.6 WATERMARKING SOFTWARE&SREVICES      Alpha-Tec: watermarking software for copyright protection and infringement tracking. Digital signature schemes in the sense used here are cryptographically based. and internet tracking. video e.  Forensic Tracking E-Commerce/Linking 2. a properly implemented digital signature gives the receiver reason to believe the message was sent by the claimed sender. A valid digital signature gives a recipient reason to believe that the message was created by a known sender such that they cannot deny sending it (authentication and non-repudiation) and that the message was not altered in transit (integrity). further. audio. embedded messages and more. Digimarc: For document verification. embedding and detecting watermarks. meaning that the signer cannot successfully claim they did not sign a message.2 DIGITAL SIGNATURE A digital signature or digital signature scheme is a mathematical scheme for demonstrating the authenticity of a digital message or document. India. partial encryption.

 A signature verifying algorithm that. and status of an electronic document as well as acknowledging informed consent and approval by a signatory. Below are some common reasons for applying a digital signature to communications: 2. Digital signatures can be used to authenticate the source of messages. The importance of high confidence in sender authenticity is especially obvious in a financial context. identity. it should be computationally infeasible to generate a valid signature for a party who does not possess the private key. or a message sent via some other cryptographic protocol. First. University of Chicago. that information may not be accurate. given a message and a private key. Secondly. public and private laws. Two main properties are required.  A signing algorithm that. The algorithm outputs the private key and a corresponding public key. For example.2.2. either accepts or rejects the message's claim to authenticity. a signature generated from a fixed message and fixed private key should verify the authenticity of that message by using the corresponding public key. given a message. contracts.1 Authentication Although messages may often include information about the entity sending a message. and Stanford are publishing electronic student transcripts with digital signatures.Digitally signed messages may be anything representable as a bitstring: examples include electronic mail. The United States Government Printing Office (GPO) publishes electronic versions of the budget. When ownership of a digital signature secret key is bound to a specific user. produces a signature. A digital signature scheme typically consists of three algorithms:  A key generation algorithm that selects a private key uniformly at random from a set of possible private keys. digital signatures can provide added assurances of the evidence to provenance. a valid signature shows that the message was sent by that user. suppose a bank's branch office . Universities including Penn State. public key and a signature.1. 2.1 Uses of Digital Signature As organizations move away from paper documents with ink signatures or authenticity stamps. and congressional bills with digital signatures.

sends instructions to the central office requesting a change in the balance of an account. an entity that has signed some information cannot at a later time deny having signed it. access to the public key only does not enable a fraudulent party to fake a valid signature. If the central office is not convinced that such a message is truly sent from an authorized source. there is no efficient way to modify a message and its signature to produce a new message with a valid signature.1. the sender and receiver of a message may have a need for confidence that the message has not been altered during transmission.2. any change in the message after signature will invalidate the signature. Although encryption hides the contents of a message. if a message is digitally signed. Similarly. By this property. is an important aspect of digital signatures. known as nonmalleable ones. 2. it may be possible to change an encrypted message without understanding it. because this is still considered to be computationally infeasible by most cryptographic hash functions (see collision resistance). but others do not.) However. acting on such a request could be a grave mistake.2 Integrity In many scenarios.1. 2.3 Non-repudiation Non-repudiation. The device signature may be in the form of  sensor pattern noise (SPN)  camera response function  Re sampling artifacts  Color filter array  Interpolation artifacts  JPEG compression  Lens aberration  sensor dust . prevent this. Furthermore.2. or more specifically non-repudiation of origin. (Some encryption algorithms.

The sensor's passbandquantum efficiency and span of the CFA's spectral responses are typically wider than the visible spectrum. Color filter array The Bayer color filter mosaic. Since sensors are made of semiconductors they obey solid-state physics. The responses of the filters do not generally correspond to the CIEcolor matching functions. so a color translation is required to convert the tristimulus values into a common. 1 blue and 1 red filter. The spectral transmittance of the CFA elements along with the demosaicing algorithm jointly determine the color rendition. or color filter mosaic (CFM). Color filters are needed because the typical photosensors detect light intensity with little or no wavelength specificity. Each two-by-two submosaic contains 2 green. thus all visible colors can be distinguished. each covering one pixel sensor. green. For example. a color filter array (CFA). is a mosaic of tiny color filters placed over the pixel sensors of an image sensor to capture color information. such that the separate filtered intensities include information about the color of light. In photography.CHAPTER 3 COLOR FILTER ARRAY 3. The color filters filter the light by wavelength range. The raw image data captured by the image sensor is then converted to a full-color image (with intensities of all three primary colors represented at each pixel) by a demosaicing algorithm which is tailored for each type of color filter. the Bayer filter (shown to the right) gives information about the intensity of light in red. and blue (RGB) wavelength regions. absolute color space. and therefore cannot separate color information. .

used in a few cameras.  List of color filter arrays Pattern (pixels) 2×2 size Image Name Description Very common RGB filter. and red sensors on top of each other. one green. This arrangement does not require a demosaicing algorithm because each pixel has information about each color. One cyan. and one magenta.The Foveon X3 sensor uses a different structure such that a pixel utilizes properties of multijunctions to stack blue. with 50% white. Traditional RGBW similar to Bayer and RGBE patterns. and one magenta. With one blue. used in a few Sony cameras. Bayer filter RGBE filter CYYM filter CYGM filter RGBW Bayer RGBW #1 2×2 2×2 2×2 2×2 RGBW #2 RGBW #3 Three example RGBW filters from Kodak. and two green.1 Manufacture of the CFA . Dick Merrill of Foveon distinguishes the approaches as "vertical color filter" for the Foveon X3 versus "lateral color filter" for the CFA. two yellow. One cyan. Bayer-like with one of the green filters modified to "emerald". (See Bayer filter#Alternatives) 4×4 2×4 3. one red. one yellow. green. used in a few cameras of Kodak.

aliasing effects. CMCR106G. the individual CFA filters are usually layers of transmissive (absorptive) organic or pigment dyes. For instance. moire patterns. and subsequently dyed."..[15] The gelatin is sectionalized. Theuwissen says "Previously. and resistance to humidity and other atmospheric stresses—is a challenging task. while other sensors have the CFA manufactured directly on the surface of the imager. Some sources indicate that the CFA can be manufactured separately and affixed after the sensor has been manufactured. CMCR106R. all single-chip color cameras are provided with an imager which has a color filter on-chip processed (Dillon. Filter materials are manufacturer specific. with the G filter being an overlap of the Y and C filters. 1982). 1978) and not as a hybrid. A few sources discuss other specific chemical substances. at best. Theuwissen makes no mention of the materials utilized in CFA manufacture. but nowadays. CMCR101G." He provides a bibliography focusing on the number. Pigment based CFAs have become the dominant option because they offer higher heat resistance and light resistance compared to dye based CFAs. . First. Ensuring that the dyes have the right mechanical properties—such as ease of application. and spatial frequencies of the absorptive filters. Nakamura said that materials for on-chip color filter arrays fall into two categories: pigment and dye. thicknesses ranging up to 1 micrometre are readily available. and optimal manufacturing processes of color filter arrays. Aoki reveals that a CYWG arrangement was used. the color filter was fabricated on a separate glass plate and glued to the CCD (Ishikawa 1981). CMCR101B. Adams et al. In either case. though solutions have been found for this problem. This makes it difficult. Color photoresists sometimes used include those with chemical monikers CMCR101R. durability. state "Several factors influence the CFA's design. and CMCR106B. attending optical properties. types.Diazonaphthoquinone (DNQ)-novolacphotoresist is one material used as the carrier for making color filters from color dyes. to fine-tune the spectral responsivities. via photolithography. There is some interference between the dyes and the ultraviolet light needed to properly expose the polymer. At least one early example of an on-chip design utilized gelatin filters (Aoki et al.

#12715.Given that the CFAs are deposited on the image sensor surface at the BEOL (back end of line.I. On the other hand.801 Koya et al. 3. the later stages of the integrated circuit manufacturing line). CAS Registry numbers. Solvent Yellow 88.808. Carl Chiulli cites the use of 5 chemicals. Image noise is an undesirable by-product of image capture that adds spurious and extraneous information. and C. mainly azo dyes and pyrazolone-diazenyl. It can be produced by the sensor and circuitry of a scanner or digital camera. . or Colour Index numbers.P. and is usually an aspect of electronic noise.I. but fail to provide chemical names. AKA Solvent Red 8.096.P. 3. three of which are C. organics would be preferred over glass. of Fuji Photo Film company. In U. where a low-temperature regime must be rigidly observed (due to the low melting temperature of the aluminum metalized "wires" and the substrate mobility of the dopants implanted within the bulk silicon).# 4. Gersteltec sells photoresists that possesses color filter properties. list some 150-200 chemical structures.3 IMAGE NOISE Image noise is random (not present in the object imaged) variation of brightness or color information in images. # 61551. # 5. some CVD silicon oxide processes are low temperature processes. Solvent Blue 36.S.2 Some pigment and dye molecules used in CFAs In U. Image noise can also originate in film grain and in the unavoidable shot noise of an ideal photon detector.. Ocean Optics has indicated that their patented dichroic filter CFA process (alternating thin films of ZnS and Cryolite) can be applied to spectroscopic CCDs.S.501.

of course.4 Types o o o o o o o Amplifier noise (Gaussian noise) Salt-and-pepper noise Shot noise Dark current noise Quantization noise (uniform noise) Read noise Anisotropic noise 3. from which a small amount of information can be derived by sophisticated processing (a noise level that would be totally unacceptable in a photograph since it would be impossible to determine even what the subject was). unwanted electrical fluctuations in signals received by AM radios caused audible acoustic noise ("static"). including that which comes from the reset noise of capacitors ("kTC noise"). inaudible. Amplifier noise is a major part of the "read noise" of an image sensor. independent at each pixel and independent of the signal intensity. that is. Gaussian. caused primarily by Johnson–Nyquist noise (thermal noise). there can be more noise in the blue channel. 3.2 Salt-and-pepper noise .Noise clearly visible in an image from a digital camera The original meaning of "noise" was and remains "unwanted signal". 3. to optical and radioastronomical images that are almost entirely noise.4.1 Amplifier noise (Gaussian noise) The standard model of amplifier noise is additive. By analogy unwanted electrical fluctuations themselves came to be known as "noise. The magnitude of image noise can range from almost imperceptible specks on a digital photograph taken in good light. In color cameras where more amplification is used in the blue color channel than in the green or red channel." Image noise is. of the constant noise level in dark areas of the image.4.

Dead pixels in an LCD monitor produce a similar. display. but non-random.Image with salt and pepper noise Fat-tail distributed or "impulsive" noise is sometimes called salt-and-pepper noise or spike noise. In addition to photon shot noise.3 Shot noise The dominant noise in the lighter parts of an image from an image sensor is typically that caused by statistical quantum fluctuations. and hot pixels appear as salt-and-pepper noise. bit errors in transmission. which is usually not very different from Gaussian. of the leakage. or if the exposure time is long enough that the hot pixel charge exceeds the linear charge capacity. and the noises at different pixels are independent of one another. Shot noise follows a Poisson distribution. 3.4. Dark current is greatest at "hot pixels" within the image sensor. leaving only the shot noise. that is. Shot noise has a root-mean-square value proportional to the square root of the image intensity. . this noise is sometimes known as "dark shot noise" or "dark-current shot noise". there can be additional shot noise from the dark leakage current in the image sensor. This noise is known as photon shot noise. etc. the noise will be more than just shot noise. or random component. It can be mostly eliminated by using dark frame subtraction and interpolating around dark/bright pixels. This type of noise can be caused by analog-to-digital converter errors. If dark-frame subtraction is not done. variation in the number of photons sensed at a given exposure level. An image containing salt-and-pepper noise will have dark pixels in bright regions and bright pixels in dark regions. The variable dark charge of normal and hot pixels can be subtracted off (using "dark frame subtraction").

3. it will be signal independent if other noise sources are big enough to cause dithering. It has an approximately uniform distribution. The read noise of a camera affects how well the image represents the actual data. 3.7 Anisotropic noise Some noise sources show up with a significant orientation in images. The efficiency of a generation center depends on its energy level.4. or if dithering is explicitly applied. Though it can be signal dependent. image sensors are sometimes subject to row noise or column noise.4.4.5 In digital cameras . with states near mid-band generating most of the dark current.4. robbing electrons of the thermal energy required to reach an intermediate state. The generation of dark current is a thermal process wherein electrons use thermal energy to hop to an intermediate state. For example. providing a path for valence electrons to sneak into the conduction band. and is present in all images taken and recorded by a camera. the most effective way to reduce dark current is to cool the CCD.[13] 3. 3.4 Dark current noise: Dark current is the result of imperfections or impurities in the depleted bulk silicon or at the siliconsilicon dioxide interface.3. from which they are emitted into the conduction band. For this reason. Calibrating the read noise allows us know more about the quality of the CCD as well as the data distortion due to the reading of images.6 Read noise Read noise is a property that is inherent to the CCD of digital cameras. adding to the signal measured in the pixel. since high read noise decreases the quality of the image. These sites introduce electronic states in the forbidden gap which act as steps between the valence and conduction bands.5 Quantization noise (uniform noise) The noise caused by quantizing the pixels of a sensed image to a number of discrete levels is known as quantization noise.

or that the f-number or focal-plane illuminance is held constant.1 second exposure. so larger sensors typically create lower noise images than smaller sensors. or effective light collection area per pixel sensor. since fewer photons are counted (shot noise) and since more amplification of the signal is necessary. That is. higher gain (ISO sensitivity). The relative effect of both read noise and shot noise increase as the exposure is reduced. In low light. the sensitivity of an imager scales roughly with the sensor area. assuming the aperture area is proportional to sensor area. . At the cost of a doubling of read noise variance (41% increase in read noise standard deviation). longer shutter speeds lead to increased salt-and-pepper noise due to photodiodeleakage currents. is the largest determinant of signal levels that determine signal-to-noise ratio and hence apparent noise levels. For images at lower signal levels (higher ISO settings).6 Effects of sensor size The size of the image sensor. for a constant f-number. when the image is scaled to the same size on screen. Banding noise. where read noise (noise floor) is significant. can be introduced through brightening shadows or through color-balance processing. 3. In the case of images bright enough to be in the shot noise limited regime.Image on the left has exposure time of >10 seconds in low light. not how this area is divided into pixels. more pixels within a given sensor area will make the image noisier if the per pixel read noise is the same. The image on the right has adequate lighting and 0. or printed at the same size. this salt-and-pepper noise can be mostly eliminated by dark frame subtraction. or both. On most cameras. the pixel count makes little difference to perceptible noise levels – the noise depends primarily on sensor area. correct exposure requires the use of long shutter speeds. similar to shadow noise. corresponding to increased ISO sensitivity.

no algorithm can make this judgment perfectly. An example shows a DSLR sensor at ISO 400 creating less noise than a point-and-shoot sensor at ISO 100. Many cameras have settings to control the aggressiveness of the in-camera noise reduction. Most noise reduction algorithms perform much more aggressive chroma noise reduction. low-contrast detail that may have characteristics similar to noise. it may be a defect in a group of pixels in the image-taking sensor (spurious and unwanted. many people find . the noise level produced by a Four Thirds sensor at ISO 800 is roughly equivalent to that produced by a full frame sensor (with roughly four times the area) at ISO 3200. But a definitive answer is not available. 3. it may be more likely to be a true feature of the image. However. but not strictly noise).For instance. and that produced by a 1/2. if it is irregular. it is likely (but not certain) to be spurious and noise. 3.7 Sensor heat Temperature can also have an effect on the amount of noise produced by an image sensor due to leakage. With this in mind.8 Image noise reduction Most algorithms for converting image sensor data to an image. whether in-camera or on a computer. so there is often a tradeoff made between noise removal and preservation of fine. which tend to use larger sensors than compacts. it is known that DSLRs will produce more noise during summer than winter. This ability to produce acceptable images at higher sensitivities is a major factor driving the adoption of DSLR cameras. and average out the former while attempting to preserve the latter. Furthermore. If this is a single pixel. involve some form of noise reduction. if it covers a few pixels in an absolutely regular shape. A simplified example of the impossibility of unambiguous noise reduction: an area of uniform red in an image might have a very small black part. There are many procedures for this.5" compact camera sensor (with roughly 1/16 the area) at ISO 100. since there is little important fine chroma detail that one risks losing. This decision can be assisted by knowing the characteristics of the source image and of human vision. but all attempt to determine whether the actual differences in pixel values constitute noise or real photographic detail.

seeking small regularities in massively random data. since its textured appearance mimics the appearance of film grain. This leads to a breakdown of image quality at higher sensitivities in two ways: noise levels increase and fine detail is smoothed out by the more aggressive noise reduction. The high sensitivity image quality of a given camera (or RAW development workflow) may depend greatly on the quality of the algorithm used for noise reduction. it is not so much a matter of noise reduction as of extracting a little information buried in a lot of noise. techniques are different. In cases of extreme noise. .luminance noise less objectionable to the eye. most camera manufacturers increase the noise reduction aggressiveness automatically at higher sensitivities. Since noise levels increase as ISO sensitivity is increased. such as astronomical images of very distant objects.

It might be affected by changes in the environment like different temperatures. Fixed pattern noise (FPN) is a general term that identifies a temporally constant lateral nonuniformity (forming a constant pattern) in an imaging system with multiple detector or picture elements (pixels). which might strongly deteriorate if additional electrical gain (and noise) is included. occurring only at certain. Sometimes pixel noise[2] as the average deviation from the array average under different illumination and temperature conditions is specified.[1] One is the DSNU (dark signal non-uniformity). imager temperature. a long exposure (integration time) emphasizes the inherent differences in pixel response so they may become a visible defect. imager gain and incident illumination. Pixel noise therefore gives a number (commonly expressed in rms) that identifies FPN in all permitted imaging conditions. material or interference with the local circuitry. . which is the offset from the average across the imaging array at a particular setting (temperature. it is not expressed in a random (uncorrelated or changing) spatial distribution. The latter can be described as the local. etc. Although FPN does not change appreciably across a series of captures. degrading the image. integration time) but no external illumination and the PRNU (photo response nonuniformity). It is characterized by the same pattern of 'hot' (brighter) and cold (darker) pixels occurring with images taken under the same illumination conditions in an imaging array. In practice. This problem arises from small differences in the individual responsibility of the sensor array (including any local post amplification stages) that might be caused by variations in the pixel size. which describes the gain or ratio between optical power on a pixel versus the electrical signal output.CHAPTER 4 FIXED PATTERN NOISE Fixed pattern noise is the term given to a particular noise pattern on digital imaging sensors often noticeable during longer exposure shots where particular pixels are susceptible to giving brighter intensities above the general background noise. exposure times. and fixed pixel locations. it may vary with integration time. pixel dependent photo response nonlinearity (PRNL) and is often simplified as a single value measured at almost saturation level to permit a linear approximation of the non-linear pixel response. The term "fixed pattern noise" usually refers to two parameters.

produced by the Optoliner. We kept the illumination level fixed at 3. 4. And it is for CCD only. since increasing the illumination level will enhance the difference in the photo-response of the pixels across the image and lead to a higher PRNU. Repeat the calculations for the different exposure times to compare the PRNU. and we also checked to ensure that the camera is focused before taking the pictures. Like DSNU.1. However. we use the camera to take multiple images of a uniform scene. 4. One of the few engineering definitions for PRNU or "photoresponsenonuniformity" is in the photonics dictionary. since the maximum value of the Opt linear device is around 4 candelas. and increasing the illumination level increases the non-uniformity of the illumination . Obtain the spatial variance of the pixel values over the entire CCD   Divide the spatial variance by the average image from (ii) to obtain the PRNU as a percentage of the actual pixel values. PRNU is the pixel variation under illumination. We expect the PRNU to increase with increasing illumination. In our measurements.2 Methods To characterize the PRNU. it is seen as the variation in pixel responsively over the CCD.1 Background Photo Response Non-Uniformity.1. while DSNU occurs as a variation in pixel responsively when the CCD is not illuminated.00 candelas since the brighter light is more easily detected by the camera. is one source of pattern noise in digital cameras. or PRNU for short. The calculation of the PRNU is as follows:  Obtain the average image over the 100 images taken:   Subtract the DSNU image from this average image to eliminate the contribution from the DSNU. We took 100 exposures each for three exposure times: 1/10. 1/4 and 1/2.5.1 PRNU (Photo Response Non-Uniformity) 4.

455) is the gamma correction factor. This colour filtering is determined by the CFA.. For most digital cameras. we chose to increase the exposure times to mimic the effect of increasing illumination levels. including . but for every pixel only therays of one colour component is passed through the CFA and subsequently converted into electronic signals by the sensor. This colour interpolation process is commonly known as demosaicking. (= 0. (1) and keeping the first two terms of the expansion where  is the denoised image and is the ensemble of the noises. . during the image acquisition process. a camera output model can be expressed as where I is the output image. primarily the JPEG format.produced by the Opt linear. The dominating component of sensor pattern noise is photo response non-uniformity (PRNU). g is the colour channel gain. these signals are stored in the camera’s memory in a customized format. The PRNU pattern noise K can then be formulated as . a colour interpolation function generates the electronic signals of the other two colour components for every pixel according to the colour intensities of the neighboring pixels. (1). shot noise. (1). Finally. and B). the PRNU can be contaminated by various types of noise introduced at different stages of the image acquisition process.s andr are random noise and is the fixed pattern noise (FPN) that is associated with every camera and can be removed by subtracting a dark frame from the image taken by the same camera. respectively. and is the input signal of the scene. stand for dark current. However. R. gamma correction and image enhancement. In Eq. and . read-out noise and quantization (lossy Compression) noise. In acquiring an image. the signal will inevitably be distorted when passing through each process and these distortions result in slight differences between the scene and the camera-captured image. the lenses let through the rays of the three colour components of the scene. after applying Taylor expansion to Eq.e. K is the zero-mean multiplicative factor responsible for the PRNU. Figure 1 demonstrates the image acquisition process. A colour photo is represented in three colour components (i. As formulated in [11]. The signals then undergo additional signal processing such as white balance. Since is the dominating term in Eq. After the conversion. G.

for each imaging device d..e. (6) is element-wise. only one of the three colours of each pixel is physically captured by the sensor while the other two are artificially interpolated by the demosaicking process. (5) is element-wise. instead of using against the reference fingerprint . (5) from a number of low-contrast images taken by device d and then the PRNU is estimated using the ML estimation procedure adopted by Chen et. (6). Again the multiplication operation in Eq. the wavelet-based denoising process (i. thus identifying the source device that has taken the image under investigation. the noise residual WI of image I under investigation is extracted using Eq. where is the mean function. we used . is the gamma correction factor . it is important that the PRNU extracted is as close to the genuine pattern noise due to the sensor as possible.is the s-th image taken by device d and Note the multiplication operation in Eq.. as suggested in [11]. Although various denoising filters can be used.is the noise residual obtained by applying a denoising filter on image I. The normalised cross-correlation is the noise residual extracted from . Given the PRNU-based approaches‟ potential in resolving device identification problem to the accuracy at individual device level. is used to compare the noise Note in Eq. where S is the number of images involved in the calculation. this inevitably introduce noise with power stronger than that of the genuine .2 Use of PRNU in Device Identification The basic idea of using the PRNU noise pattern in device identification can be described as follows.. has been reported as effective in producing good results. al. 1) First. Since for most cameras. the discrete wavelet transform followed by a Wiener filtering operation). 2) Secondly.e. (5) and compared against the reference PRNU Kd of each device d available to the investigator in the hope that it will match one of the reference fingerprints. 4. i. the noise residual patterns are extracted using Eq.

Addressing this shortcoming is the motivation of this work. The algorithm should have the following traits:  Avoidance of the introduction of false color artifacts. a full set of color triples) from the spatially under sampled color channels output from the CFA. as a result. (3) and (5) as a function of the noise residual W (i. In this work. as mentioned earlier that I is to apply the discrete wavelet transform followed by a Wiener filtering operation directly to the entire image I without differentiating physical components from artificial components and. zippering (abrupt unnatural changes of intensity over a number of neighboring pixels) and purple fringing  Maximum preservation of the image resolution .. so demosaicing is part of the processing pipeline required to render these images into a viewable format. Most modern digital cameras acquire images using a single image sensor overlaid with a CFA.PRNU.e. (3) and (4) that the accuracy of both PRNU K and noise residual W depends on the denoising operation applied to I in obtaining the most common method of obtaining .e. Many modern digital cameras can save images in a raw format allowing the user to demosaic it using software. Because the PRNU is formulated in Eq. we present some experiments on device identification and image content integrity verification to validate the proposed PRNU extractionformula. The aim of a demosaicing algorithm is to reconstruct a full color image (i. rather than using the camera's built-in firmware.3 DEMOSAICING A demosaicing (also de-mosaicing or demosaicking) algorithm is a digital image process used to reconstruct a full color image from the incomplete color samples output from an image sensor overlaid with a color filter array (CFA). It is also known as CFA interpolation or color reconstruction. such as chromatic aliases. Eq. Section V concludes this work. We can see from Eq. In Section IV. 4. PRNU and noise residual. we will look at the impact of demosaicking on PRNU fidelity in Section II and propose an improved formula for extracting PRNU in Section III. However. (2). interchangeably whenever there is no need to differentiate them. allowing the interpolation noise in the artificial components to contaminate the real PRNU in the physical components. (4)). in the rest of the work we will use the two terms.

. hence the accuracy of colour interpolation result is also content-dependent. in inhomogeneous areas. (4) and (5) based on the definition of the output image model in Eq. As a result the interpolation noise gets diffused from the artificial components into the physical ones. However. However. each coefficient of the wavelet transform used in the noise residual extraction process involves multiple pixels and thus both artificial and physical components. the existing method for extracting PRNU as formulated in Eq. e. we call the colour components physically captured by the sensor as physical colours and the ones artificially interpolated by the demosaicking function as artificial colours. in one noise residual extraction process. To extract the PRNU using Eq. Therefore it is desirable to devise a noise residual extraction method that can . Due to the fact that demosaicking is a key deterministic process that affects the quality of colour images taken by many digital devices. 4. The mathematics here is subject to individual implementation. thus the interpolation noise is also more significant. in a homogeneous area. Conversely. Low computational complexity for fast processing or efficient incamera hardware implementation  Amenability to analysis for accurate noise reduction To reconstruct a full color image from the data collected by the color filtering array. For example. because of the low variation of the colour intensities of neighbouring pixels. demosaicking has been rigorously investigated. and is called demosaicing. For example. the interpolation function can more accurately generate artificial components. edge-adaptive or non-adaptive.4 DEMOSAICKING IMPACT ON PRNU FIDELITY In this work. This indicates that the PRNU in physical colour components is more reliable than that in the artificial components. a form of interpolation is needed to fill in the blanks. (1) does not take this into account. in the red colour component/plane of an image taken by a camera with a Bayer CFA. the discrete wavelet transform followed by a Wiener filtering operation is applied. only one fourth of the pixels‟ red colour are physical and for each pixel with physical red colour all its 8 neighbours‟ red colours are artificial. The grouping process is usually content-dependent. (4) and (5). (4) is that it involves the whole image plane.g. the colour variation between neighbouring pixels is greater. Most demosaicking approaches group the missing colours before applying an interpolation function. which contains both artificial and physical components. The main problem inherent to Eq. When wavelet transform is applied during the noise residual extraction process the interpolation noise residing in the artificial components propagates into the physical components.

prevent the artificial components from contaminating the reliable PRNU residing in the physical components with the interpolation noise. .

Let be an interpolation matrix with 2N+1 × 2N+1 coefficients and be a X × Y-pixel input signal from the scene consisting of three colour components. (7) means that if the colour component c is the same as the colour that the CFA pattern F allows to pass.CHAPTER 5 CD-PRNU (Color Decoupled Photo Response Non-Uniformity) 5.e . For each colour component of a pixel . R (red). Otherwise. The other two colour components are to be determined by the demosaicking process. the image output model of Eq. the second part of Eq.1 FORMULATION OF COLOUR DECOUPLED PRNU (CD-PRNU) In this section. That is to say that for each pixel . F. we will discuss the formulation and extraction of CD-PRNU. (7). can be determined according to The first part of Eq. According to Eq.2 Mathematical Model of CD-PRNU A generic demosaicking process is to convolve an interpolation matrix with an image block of the same size centred at the pixel where the artificial colour is to be calculated. only one of the three colour components takes a value physically captured by the sensor and this colour is determined by the colour configuration of the CFA pattern F. (7) is artificially applied to calculate the colour. i. G (green) and B (blue) before colour interpolation. 5. without prior knowledge about the CFA. Although the 2×2 Bayer CFA is the most common CFA pattern. then no demosaicking is needed because c has been physically captured by the sensor. (1) proposed in can be reformulated as . First. except that it is a 2 × 2 square array. to make the proposed CD-PRNU versatile and applicable to cameras adopting different CFA patterns. a mathematical model for the CD-PRNUis derived and then an extraction algorithm is proposed to extract the noise residual that is to be used for estimating the final CD-PRNU. we makes no assumption about the CFA pattern.

which is actually the interpolation noise.B. (9) suggests that in the artificial components. That is where is a low-passed filtered version of the artificial components and is the corresponding ―sensor pattern noise‖. Hence if the physical and artificial colour components can be separated / decoupled. We will discuss how the physical and artificial is the interpolation noise extracted from colour components can be decoupled in simple manner without a priori knowledge about the CFA pattern in Section III. the PRNU is actually the interpolation noise P while. (5) to extract the reference interpolation noise d from S low-variation images taken by d such that for a particular device where is the artificial colour components of the s-th low-contrast image taken by device d and . Eq. in the physicalcomponents. P can be extracted in the same way as the sensor pattern noise K is extracted (i. 5.3 CD-PRNU Extraction Algorithm .e.. (3)).Eq. We can also use the same ML estimate as in Eq. the PRNU remains unaffected by the interpolation noise. (9) that the physical components and artificial components have similar mathematical expression. It can also be seen from Eq.

. from the physical and artificial components if the CFA is known. For each colour channel. to extract the CD-PRNU. . respectively. We know that. . In this work. (11) is applied. Eq. (10) and (11). manufacturers usually do not provide information about the CFA used by their cameras. for each pixel of I. To estimate the CD-PRNU Pd of a particular device d and use it as the reference signature of d. Finally the CD-PRNU Wc of each colour channel c is formed by combining the four sub-noise residuals such that where. we first separate the three colour channels of a colour image I of pixels. only one of the three colour components is physical and the other two are artificial. we do not know (actually we do not have to know) which pixels carry the colour captured physically by the hardware and which are not. such that vertical dimensions to get four sub-images. so the second step is. Note that Algorithm 1 is for extracting the noise residual pattern W from an image I. Therefore.According to Eq. (4) is then used to obtain noise residual from each sub-images . we can prevent the artificial components from contaminating the physical components during the DWT process. several methods have been proposed to estimate the CFA. and mod is the modulo operation. these methods have to exhaust all of the possible CFA patterns in order to infer/estimate the „real‟/optimal CFA. Most CFA patterns are of 2 × 2 elements and are periodically mapped to the sensors. Unfortunately. exhaustive search is by no means acceptable. Eq. However. we can extract the sensor pattern noise and interpolation noise. we perform a 2:1 down-sampling across both horizontal and . But by decomposing into four sub-images. we know that each of the four sub-images either contains only the physical colour or only the artificial colours. for each channel . The framework of the colour decoupled noise residual extraction process is shown in Figure 2 and the procedures are listed in Algorithm 1. By decoupling the physical and artificial colour components in this fashion before extracting the noise residual. However. without knowing the CFA pattern used by the manufacturer.

.

we carry out experiments on source camera identification and image content integrity verification to validate the feasibility of the proposed CD-PRNU in a comparative manner.e. 5. The reference PRNU (i.5.1. Table 1. The six cameras are listed in Table1.5 EXPERIMENTAL RESULTS In this section. ) of each camera Ci is generated by taking the weighted average of the PRNUs extracted from 30 photos of blue sky according to Eq.5. we need clean PRNUs (which appear as high frequency bands of images) as device .4 Algorithm 1. Source Camera Identification We have carried out source camera identification tests on 300 2048×1536-pixel photos of natural scenes taken by six cameras(C1 to C6). Noise residual extraction algorithm Input: original image I Output: colour decoupled noise residual W Noise residual extraction algorithm 5. (11). For device identification purpose. Cameras used in the experiments. each responsible for 50.

except for a few cases when images are of full size (1536 × 2048 pixels) and the identification rates are close or equal to 100% (1. thus giving better chance of extracting clean PRNU.e. In practice. because the smaller the images. . other images with low-variation scenes (i. a camera is identified as the source camera. C1. the less data is available. of the image I under investigation. Therefore. Source camera identification requires similarity comparisons among PRNUs (CD-PRNUs) and therefore the feasibility of the chosen similarity metrics is important. The reason blue-sky images are chosen in this work is because blue sky contains less scene details (high frequency signal). normalised cross-correlation formulated as in Eq. Table 2 lists the identification rates.fingerprints for comparison against the PRNU extracted from individual images under investigation. therefore identification results become less reliable. C4. Actually. we also compare the performance of the proposed CD-PRNU against that of the traditional PRNU [11] when they are applied to blocks of 5 different sizes cropped from the centre of the fullsized PRNU (CD-PRNU). Because PRNU is often used in content integrity verification. Individually speaking.0000). C5 and C6 perform significantly better when CD-PRNU is used in all cases. The purpose of this experiment is to demonstrate the capability of the proposed CD-PRNU in dealing with the colour interpolation noise. PRNU performs equally well as CD-PRNU when the image size is 192 × 256 pixels and slightly outperforms CD-PRNU when the block size is 48 × 64 pixels. Our empirical experience suggests that an average of 20 blue sky images is accurate enough. Another reason is that. so geometrical transformations will not be applied in order to prevent biased evaluation from happening. (6) will be used to measure the similarity between PRNUs (CD-PRNUs). which has been proved to be a more stable detection statistics than normalised cross-correlation when applied to the scenarios in which the images of interest may have undergone geometrical manipulations. Therefore. Generally speaking.. C3. its reference PRNU (or CD-PRNU) is most similar to the PRNU (or CD-PRNU). Fridrich suggested the use of the Peak to Correlation Energy (PCE) measure in [15]. where smaller image blocks have to be analysed. We suspect that the reason C2 does not perform as expected is because the CFA pattern is not a 2 × 2 square array as we have assumed. Taking the average of the PRNUs from 30 blue sky images is to further reduce variation. the key point is about demonstrating the different performance of the traditional PRNU and the proposed CD-PRNU. if out of the six reference PRNUs (or CDPRNUs). in the following experiments. the normalised cross-correlation has to be greater than a specified threshold for a camera to be identified as the source camera. scenes without significant details) can be used instead. in this experiment. However. WI. For C2. such as rotation or scaling.

This has been graphically presented in Figure 3(a). Performance comparison of source camera identification a) Overall identification rates when CD-PRNU and PRNU are used as fingerprint In Figure 3(b). . as listed in the Total column of Table 2. Figure 3. We can see that the CD-PRNU outperforms the PRNU because at all fixed False Positive rate the CDPRNU‟s True Positive rate are always higher than that of the PRNU. a ROC curve of the performance of PRNU and CD-PRNU are demonstrated. we can see that CD-PRNU still outperforms PRNU significantly.when the statistics of the six cameras are pooled together.

Although extracting CD-PRNU requires down-sampling and up-sampling.Figure 3. The amount of data processed during the extraction of PRNU and CD-PRNU is the same. it takes 0. these two operations are trivial and only incur negligible increase of time complexity.567 seconds to calculate the similarity between a pair of CD-PRNUs of the same size. . Performance comparison of source camera identification b) Overall ROC curve when CDPRNU and PRNU are used as fingerprint For a system with a Pentium Core II 1.526 seconds to compute the similarity between the PRNUs of two images of 2048 × 1536 pixels and 0.3G CPU and 3 GB RAM.

. 5.2 in Figure 4(b) to create the forged Image I.Table 2. and pasted it at approximately the same location in Image I.2 Content Integrity Verification We also carried out the following three content integrity verification experiments on 640 × 480-pixel images. Source camera identification rates using traditional PRNU and proposed CD-PRNU.5.3 as shown in Figure 4(c). The images in Figure 4(a) and (b) are taken by Olympus C730. we copied a 160 × 390-pixel area from Image I.1 in Figure 4(a).  In the first experiment.

to create the forged Image III. which is taken by Olympus C730. II. we cropped an 80 × 100-pixel area from Image II. The original image. we slid a 128 × 128-pixel window across the PRNU extracted from the image under investigation and another window of the same size across the reference PRNU of the cameras that have taken images I. In Chen’s method [11]. (a) Original Image II. The images in Figure 5(a) and (b) are also taken by the same camera.2 in Figure 6(b).3  In the third experiment.1 in Figure 6(a) taken by Canon Power Shot A400.1 in Figure 5(a).3 in Figure 5(c).2 (c) Forged Image I. we cropped a 60 × 80-pixel area from Image III.3 in Figure 6(c). Figure 5.2 and III. which covers the face of the person. source image and forged images for the content verification experiments.2 (c) Forged Image III. (a) Original Image III.3 To detect the manipulated areas. Figure 6.1 (b) Original Image III. The original image. source image and forged images for the content verification experiments.1 (b) Original Image II.3  In the second experiment. the windows are .1 (b) Original Image I.2.Figure 4. pasted it at the area where the face of another person is in Image III.2 (c) Forged Image II.2 in Figure 5(b) to create the forged Image II. source image and forged images for the content verification experiments. which covers the face of the person. (a) Original Image I. pasted it at the area where the face of another person is in Image II. The original image.2.

(6). the cross-correlation of the PRNU patterns inside the two windows at the same location was calculated according to Eq. If the cross-correlation is lower than a predetermined threshold t. . in our experiment. true positive (TP). Let B be an arbitrary block and M(B) and Md(B) be defined as TP.0 and use the four metrics. Number of manipulated and non-manipulated areas in each image (unit: block). we use various thresholds defined as to analyze the performance of PRNU and CD-PRNU. true negative (TN) and false negative (FN) to measure the performance of integrity verifications based on PRNU and CD-PRNU. the sliding step/displacement is set to 5 pixels in order to reduce the computational load without sacrificing the accuracy of the integrity verification. we can evaluate the integrity verification performance across a wide range of correlation thresholds T(t). the cross-follows the Generalized Gaussian (GG) distribution. TN = |{B | M(B) = 0 and Md(B) = 0}|. and lower FP and FN indicate better performance. To decide whether a block centered at the window superposed on the image has been manipulated or not. Table 3.moved a pixel at a time. we will obtain lower TP and FP. As t grows. FP = |{B | M(B) = 0 and Md(B) = 1}| and FN = |{B | M(B) = 1 and Md(B) = 0}|. In the following experiments we will allow t to vary independently in the range from 0. TN and FN are defined as TP = |{B | M(B) = 1 and Md(B) = 1}|. which incurs a high computational load. respectively. As discussed in [11]. false positive (FP). this method is not accurate at the pixel level [11]. and T(t) is the threshold.0 to 3. Higher TP and TN. while higher TN and FN. Moreover. where and are the mean and standard deviation of the correlations distribution. the block in the centre of the window is deemed as manipulated. Therefore. By varying the value of t. therefore. Table 3 lists the number of manipulated and non-manipulated blocks of 5 × 5 pixels in the forged images. FP.

the performance of the filtering / dilation operation strongly depends on parameter setting and hence many experiments must be run to obtain the best parameters for filtering. ―the block dimensions impose a lower bound on the size of tampered regions that our algorithm can identify. which is marked along the horizontal axis of the ROC curve. which is marked vertically. Therefore a ROCcurve of TP rate with respect to FP rate can be used to evaluate the overall performance of the PRNU and CD-PRNU. 5. Thus. an algorithm with better performance will have a higher true positive rate (). TN. therefore the PRNU pattern noises in the two areas are almost the same. We can see from Figure 7(a) and 7(b) that CD-PRNU generally achieves higher TP and TN while maintaining lower FP and FN.3 is illustrated as Figure 8. a low threshold also results in the situation where more authentic blocks are mistakenly detected as manipulated. As a result. This is because the area we copied from the source image I. leading to higher TP.2. we remove all simply connected tampered regions from Z that contain less than 64×64 pixels (one quarter of the number of pixels in the block)‖. However. FP and FN. The final authentication result is a image with the dilated areas highlighted as the tampered areas. Chen applies erosion and dilation operations with a square kernel in order to filter small areas identified as tampered with. It is clear that the ROC curve of the PRNU-based scheme mostly overlaps with that of Random Guess. In order to simplify the comparison and to obtain a fair result. we use the raw data without any filtering to calculate the TP. the ROC is formulated as At the same false positive rate . TN. the experiments on III.According to Chen‟s predication. A lower correlation (similarity) allows the algorithm to detect more manipulated blocks.5. As a result.1 is at approximately the same location as the original area in image I. However. the scheme cannot detect the manipulated area based on PRNU.1 Experiment on Image I.3 Figure 7 shows the performance of the PRNU and CD-PRNU in terms of TP. By .3 across a range of correlation threshold T(t).2.3 demonstrate that CD-PRNU-based method significantly outperforms the PRNU-based method when the tampered area is about one quarter of the sliding window. The ROC curves for the integrity verification experiments on image I. FP and FN when authentication is carried out on image I. Let α be the number of manipulated blocks and β be the number of authentic blocks. giving rise to a higher FP. which means the authentication result is generally as unreliable as that of a random guess.

3. Authentication results on image I.0 to 3.0.2.5. Figure 8. Figure 7. which means that by using CD-PRNU manipulated blocks can be detected more reliably.contrast.3: Integrity verification performance of the PRNU and CD-PRNU in terms of a) TP. 5. with t varying from 0. b) TN.2 Experiment on Image II. across a range of correlation threshold T(t). The ROC curve of Truth Positive Rate with respect to False Positive Rate of PRNU and CD-PRNU when authentication is performed on image I.3 . the CD-PRNU-based scheme results in a curve much higher than the PRNU-based method.

b) TN.When verifying the integrity of image II. c) FP and d) FN across a range of correlation threshold T(t).3: Integrity verification performance of the PRNU and CD-PRNU in terms of a) TP. with t varying from 0.3. again indicate its superiority to PRNU. CD-PRNU‟s consistently higher TP and lower FN.0 to 3. Authentication results on image II. However.0. as shown in Figure 9(a) and 9(d). both PRNU and CD-PRNU can effectively detect the manipulated blocks as their ROC curves have suggested in Figure 10. Figure 10 also shows that the ROC curve of CD-PRNU is still slightly higher than that of PRNU. indicating a slightly better performance of CD-PRNU. Figure 9. mixed performance in terms of TN and FP can be seen in Figure 9(b) and 9(c). Albeit their mixed performance in terms of TN and FP. .

The ROC curve of Truth Positive Rate with respect to False Positive Rate of PRNU and CD-PRNU when authentication is performed on image II.2.3 Experiment on Image III. the method based on PRNU can perform no better than a random guess. the CD-PRNU-based method can identify smaller areas. the manipulated blocks can be effectively detected by the CD-PRNU-based scheme because the areas in question are from two images taken by different cameras and thus contain different interpolation noise.3 When authenticating III. which is only about one quarter of the sliding window (128 × 128 pixels).1. and therefore areas smaller than this should be filtered in order to remove the falsely identified noise. the PRNU cannot correctly detect any manipulated blocks (i. although the performance of PRNU and CD-PRNU in terms of TN and FP are mixed.3.5. as can be seen in Figure 11(b) and 11(c). . respectively. By contrast. Chen predicated in that one quarter of the sliding window is the lower bound on the size of tampered regions that our algorithm can identify. Since the tampered area is 60 × 80 pixels. approximately one quarter of the window.Figure 10.e. As a result. CD-PRNU‟s significantly better performance in terms of TP and lower FN can still be seen again in Figure 11(a) and 11(d). The experiment result on III. This poor performance is also reflected in the PRNU’s ROC curve in Figure 12 and is due to the fact that he manipulated area is too small (60 × 80 pixels).3 conforms to Chen’s observation. as demonstrated in Figure 11(a). When the threshold t is higher than 1. 5.3.

Authentication results on image III.3: Integrity verification performance of the PRNU and CD-PRNU in terms of a) TP.0 to 3. b) TN.Figure 11. .0. c) FP and d) FN across a range of correlation threshold T(t). with t varying from 0.

Sign up to vote on this title
UsefulNot useful

Master Your Semester with Scribd & The New York Times

Special offer: Get 4 months of Scribd and The New York Times for just $1.87 per week!

Master Your Semester with a Special Offer from Scribd & The New York Times