Discrete Wavelet Transform (DWT)

Presented by
Sharon Shen
UMBC
Overview
Introduction to Video/Image Compression
DWT Concepts
Compression algorithms using DWT
DWT vs. DCT
DWT Drawbacks
Future image compression standard
References

Need for Compression
Transmission and storage of uncompressed
video would be extremely costly and
impractical.
Frame with 352x288 contains 202,752 bytes of information
Recoding of uncompressed version of this video at 15 frames
per second would require 3 MB. One minute180 MB
storage. One 24-hour day262 GB
Using compression, 15 frames/second for 24 hour1.4 GB,
187 days of video could be stored using the same disk space
that uncompressed video would use in one day
Principles of Compression
Spatial Correlation
Redundancy among neighboring pixels
Spectral Correlation
Redundancy among different color planes
Temporal Correlation
Redundancy between adjacent frames in a
sequence of image

Classification of Compression
Lossless vs. Lossy Compression
Lossless
Digitally identical to the original image
Only achieve a modest amount of compression
Lossy
Discards components of the signal that are known to be
redundant
Signal is therefore changed from input
Achieving much higher compression under normal viewing
conditions no visible loss is perceived (visually lossless)
Predictive vs. Transform coding
Classification of Compression
Predictive coding
Information already received (in transmission) is used
to predict future values
Difference between predicted and actual is stored
Easily implemented in spatial (image) domain
Example: Differential Pulse Code Modulation(DPCM)


Classification of Compression
Transform Coding
Transform signal from spatial domain to other space
using a well-known transform
Encode signal in new domain (by string coefficients)
Higher compression, in general than predictive, but
requires more computation (apply quantization)
Subband Coding
Split the frequency band of a signal in various
subbands

Classification of Compression
Subband Coding (cont.)
The filters used in subband coding are known as
quadrature mirror filter(QMF)
Use octave tree decomposition of an image data into
various frequency subbands.
The output of each decimated subbands quantized
and encoded separately
Discrete Wavelet Transform
The wavelet transform (WT) has gained widespread
acceptance in signal processing and image compression.
Because of their inherent multi-resolution nature,
wavelet-coding schemes are especially suitable for
applications where scalability and tolerable degradation
are important
Recently the JPEG committee has released its new image
coding standard, JPEG-2000, which has been based
upon DWT.

Discrete Wavelet Transform
Wavelet transform decomposes a signal into a set of
basis functions.
These basis functions are called wavelets
Wavelets are obtained from a single prototype wavelet
y(t) called mother wavelet by dilations and shifting:

(1)


where a is the scaling parameter and b is the shifting
parameter
) (
1
) (
,
a
b t
a
t
b a

=
Discrete Wavelet Transform
Theory of WT
The wavelet transform is computed separately for different
segments of the time-domain signal at different frequencies.
Multi-resolution analysis: analyzes the signal at different
frequencies giving different resolutions
MRA is designed to give good time resolution and poor
frequency resolution at high frequencies and good frequency
resolution and poor time resolution at low frequencies
Good for signal having high frequency components for short
durations and low frequency components for long duration.e.g.
images and video frames
Discrete Wavelet Transform
Theory of WT (cont.)
Wavelet transform decomposes a signal into a set of basis
functions.
These basis functions are called wavelets
Wavelets are obtained from a single prototype wavelet y(t) called
mother wavelet by dilations and shifting:

(1)


where a is the scaling parameter and b is the shifting
parameter
) (
1
) (
,
a
b t
a
t
b a

=
Discrete Wavelet Transform
The 1-D wavelet transform is given by :


Discrete Wavelet Transform
The inverse 1-D wavelet transform is given
by:


Discrete Wavelet Transform
Discrete wavelet transform (DWT), which transforms a discrete
time signal to a discrete wavelet representation.
it converts an input series x
0
, x
1
, ..x
m
, into one high-pass wavelet
coefficient series and one low-pass wavelet coefficient series (of
length n/2 each) given by:
Discrete Wavelet Transform
where s
m
(Z) and t
m
(Z) are called wavelet filters, K is the length of the
filter, and i=0, ..., [n/2]-1.
In practice, such transformation will be applied recursively on the
low-pass series until the desired number of iterations is reached.
Discrete Wavelet Transform
Lifting schema of DWT has been recognized as a faster
approach
The basic principle is to factorize the polyphase
matrix of a wavelet filter into a sequence of
alternating upper and lower triangular matrices and
a diagonal matrix .
This leads to the wavelet implementation by means of
banded-matrix multiplications
Discrete Wavelet Transform
Two Lifting schema:
Discrete Wavelet Transform






where s
i
(z) (primary lifting steps) and t
i
(z) (dual lifting steps) are filters
and K is a constant.
As this factorization is not unique, several {s
i
(z)}, {t
i
(z)} and K are
admissible.


Discrete Wavelet Transform
2-D DWT for Image


Discrete Wavelet Transform
Discrete Wavelet Transform
2-D DWT for Image


Discrete Wavelet Transform
Integer DWT
A more efficient approach to lossless compression
Whose coefficients are exactly represented by finite
precision numbers
Allows for truly lossless encoding
IWT can be computed starting from any real valued
wavelet filter by means of a straightforward modification
of the lifting schema
Be able to reduce the number of bits for the sample
storage (memories, registers and etc.) and to use simpler
filtering units.
Discrete Wavelet Transform
Integer DWT (cont.)
Discrete Wavelet Transform
Compression algorithms using DWT
Embedded zero-tree (EZW)
Use DWT for the decomposition of an image at each level
Scans wavelet coefficients subband by subband in a zigzag
manner
Set partitioning in hierarchical trees (SPHIT)
Highly refined version of EZW
Perform better at higher compression ratio for a wide
variety of images than EZW

Discrete Wavelet Transform
Compression algorithms using DWT (cont.)
Zero-tree entropy (ZTE)
Quantized wavelet coefficients into wavelet trees to reduce
the number of bits required to represent those trees
Quantization is explicit instead of implicit, make it possible to
adjust the quantization according to where the transform
coefficient lies and what it represents in the frame
Coefficient scanning, tree growing, and coding are done in
one pass
Coefficient scanning is a depth first traversal of each tree
Discrete Wavelet Transform

DWT vs. DCT
Discrete Wavelet Transform
Disadvantages of DCT
Only spatial correlation of the pixels inside the single
2-D block is considered and the correlation from the
pixels of the neighboring blocks is neglected
Impossible to completely decorrelate the blocks at
their boundaries using DCT
Undesirable blocking artifacts affect the
reconstructed images or video frames. (high
compression ratios or very low bit rates)
Discrete Wavelet Transform
Disadvantages of DCT(cont.)
Scaling as add-onadditional effort
DCT function is fixedcan not be adapted to source
data
Does not perform efficiently for binary images (fax or
pictures of fingerprints) characterized by large
periods of constant amplitude (low spatial
frequencies), followed by brief periods of sharp
transitions

Discrete Wavelet Transform
Advantages of DWT over DCT
No need to divide the input coding into non-overlapping 2-D
blocks, it has higher compression ratios avoid blocking
artifacts.
Allows good localization both in time and spatial frequency
domain.
Transformation of the whole image introduces inherent
scaling
Better identification of which data is relevant to human
perception higher compression ratio

Discrete Wavelet Transform
Advantages of DWT over DCT (cont.)
Higher flexibility: Wavelet function can be freely chosen
No need to divide the input coding into non-overlapping 2-
D blocks, it has higher compression ratios avoid blocking
artifacts.
Transformation of the whole image introduces inherent
scaling
Better identification of which data is relevant to human
perception higher compression ratio (64:1 vs. 500:1)

Discrete Wavelet Transform
Performance
Peak Signal to Noise ratio used to be a measure of image
quality
The PSNR between two images having 8 bits per pixel or sample
in terms of decibels (dBs) is given by:
PSNR = 10 log
10

mean square error (MSE)
Generally when PSNR is 40 dB or greater, then the original and
the reconstructed images are virtually indistinguishable by
human observers
|
|
.
|

\
|
MSE
2
255
Discrete Wavelet Transform
Improvement in PSNR using DWT-JEPG over DCT-JEPG at S = 4
PSNR Difference vs. Bit rate
0
0.5
1
1.5
2
2.5
0.2 0.3 0.4 0.5 0.6
Bit rate (bps)
P
S
N
R

d
i
f
f
.

(
d
B
s
)
DWT-JPEG
Discrete Wavelet Transform
Discrete Wavelet Transform
images.

Compression ratios used for 8-bit 512x512
Lena image.

8

16

32

64

128

PSNR (dBs) performance of baseline JPEG
using on Lena image.

38.00

35.50

31.70

22.00

2.00


PSNR (dBs) performance of Zero-tree
coding using arithmetic coding on Lena
image.

39.80

37.00

34.50

33.00

29.90

PSNR (dBs) performance of bi-orthogonal
filter bank using VLC on Lena image.

35.00

34.00

32.50

28.20

26.90

PSNR (dBs) performance of bi-orthogonal
filter bank using FLC on Lena image.

33.90

32.80

31.70

27.70

26.20

PSNR (dBs) performance of W6 filter bank
using VLC on Lena image.

33.60

32.00

30.90

27.00

25.90

PSNR (dBs) performance of W6 filter bank
using FLC on Lena image.

29.60

29.00

27.50

25.00

23.90

Comparison of image compression results using DCT and DWT
Discrete Wavelet Transform
Visual Comparison

(a) (b) (c)
(a) Original Image256x256Pixels, 24-BitRGB (b) JPEG (DCT)
Compressed with compression ratio 43:1(c) JPEG2000 (DWT)
Compressed with compression ratio 43:1
Discrete Wavelet Transform
Implementation Complexity
The complexity of calculating wavelet transform depends on the
length of the wavelet filters, which is at least one multiplication
per coefficient.
EZW, SPHIT use floating-point demands longer data length
which increase the cost of computation
Lifting schemea new method compute DWT using integer
arithmetic
DWT has been implemented in hardware such as ASIC and
FPGA
Discrete Wavelet Transform
Resources of the ASIC used and data processing rates for DCT
and DWT encoders



Type of
encoders
using
ASIC

No. of
Logic
gates of
the ASIC
used

Amount of on
chip RAM
used by the
encoders

Data
processing
rates of the
encoders
using ASIC

DCT

34000

128 byte

210 MSa/sec

DWT

55000

55 kbyte

150 MSa/sec

Discrete Wavelet Transform
Number of logic gates


No. of logic gates used vs. Compression
technique
0
20000
40000
60000
DCT DWT
Compression technique
N
o
.

o
f

l
o
g
i
c

g
a
t
e
s

u
s
e
d
No. of logic gates
used
Discrete Wavelet Transform
Processing Rate


Discrete Wavelet Transform
Disadvantages of DWT
The cost of computing DWT as compared to DCT
may be higher.
The use of larger DWT basis functions or wavelet
filters produces blurring and ringing noise near edge
regions in images or video frames
Longer compression time
Lower quality than JPEG at low compression rates




Discrete Wavelet Transform
Future video/image compression
Improved low bit-rate compression performance
Improved lossless and lossy compression
Improved continuous-tone and bi-level compression
Be able to compress large images
Use single decompression architecture
Transmission in noisy environments
Robustness to bit-errors
Progressive transmission by pixel accuracy and resolution
Protective image security



Discrete Wavelet Transform
References
http://www.ii.metu.edu.tr/em2003/EM2003_presentations/DSD/b
enderli.pdf
http://www.etro.vub.ac.be/Members/munteanu.adrian/_private/
Conferences/WaveletLosslessCompression_IWSSIP1998.pdf
http://www.vlsi.ee.upatras.gr/~sklavos/Papers02/DSP02_JPEG2
00.pdf
http://www.vlsilab.polito.it/Articles/mwscas00.pdf
M. Martina, G. Masera , A novel VLSI architecture for integer
wavelet transform via lifting scheme, Internal report, VLSI Lab.,
Politecnico diTor i no, Jan. 2000, unpublished.
http://www.ee.vt.edu/~ha/research/publications/islped01.pdf




Discrete Wavelet Transform

THANK YOU !

Q & A