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 minute180 MB
storage. One 24-hour day262 GB
 Using compression, 15 frames/second for 24 hour1.4 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) 1 t b
 a ,b (t )  ( )
a a
where a is the scaling parameter and b is the shifting
parameter
 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)
1 t b
 a ,b (t )  ( )
a a
where a is the scaling parameter and b is the shifting
parameter
 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 x0, x1, ..xm, 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 sm(Z) and tm(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 si(z) (primary lifting steps) and ti(z) (dual lifting steps) are filters
and K is a constant.
As this factorization is not unique, several {si(z)}, {ti(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-onadditional effort
 DCT function is fixedcan 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:
 255 
2
 PSNR = 10 log10  MSE 
 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
 Discrete Wavelet Transform

 Improvement in PSNR using DWT-JEPG over DCT-JEPG at S = 4

PSNR Difference vs. Bit rate

2.5
PSNR diff. (dBs)

1.5
DWT-JPEG
1

0.5

0
0.2 0.3 0.4 0.5 0.6
Bit rate (bps)
Discrete Wavelet Transform
images.

Discrete Wavelet Transform

Compression ratios used for 8-bit 512x512 8 16 32 64 128

Lena image.

PSNR (dBs) performance of baseline JPEG 38.00 35.50 31.70 22.00 2.00
using on Lena image.  

PSNR (dBs) performance of Zero-tree 39.80 37.00 34.50 33.00 29.90

coding using arithmetic coding on Lena
image.
PSNR (dBs) performance of bi-orthogonal 35.00 34.00 32.50 28.20 26.90
filter bank using VLC on Lena image.

PSNR (dBs) performance of bi-orthogonal 33.90 32.80 31.70 27.70 26.20

filter bank using FLC on Lena image.

PSNR (dBs) performance of W6 filter bank 33.60 32.00 30.90 27.00 25.90
using VLC on Lena image.

PSNR (dBs) performance of W6 filter bank 29.60 29.00 27.50 25.00 23.90
using FLC on Lena image.

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 schemea 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 No. of Amount of on Data

encoders Logic chip RAM processing
using gates of used by the rates of the
ASIC the ASIC encoders encoders
used 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. Com pression

technique
No. o f logic gates

60000

40000
used

No. of logic gates

20000 used

0
DCT DWT
Com pression technique
 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/
benderli.pdf
 http://www.etro.vub.ac.be/Members/munteanu.adrian
/_private/Conferences/WaveletLosslessCompression
_IWSSIP1998.pdf
 http://www.vlsi.ee.upatras.gr/~sklavos
/Papers02/DSP02_JPEG200.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