Embedded Zerotree Wavelet

- An Image Coding Algorithm

Shufang Wu [Link]
vswu@[Link] Friday, June 14, 2002

Embedded Zerotree Wavelet

22 - 1

Agenda
Overview Discrete Wavelet Transform Zerotree Coding of Wavelet Coefficients Successive-Approximation Quantization (SAQ) Adaptive Arithmetic Coding Relationship to Other Coding Algorithms A Simple Example Experimental Results Conclusion References Q&A
22 - 2

Embedded Zerotree Wavelet

Overview (2-1)
Two-fold problem
Obtaining best image quality for a given bit rate Accomplishing this task in an embedded fashion

What is Embedded Zerotree Wavelet (EZW) ?

An embedded coding algorithm 2 properties, 4 features and 2 advantages (next page)

What is Embedded Coding?

Representing a sequence of binary decisions that distinguish an image from the null image Similar in spirit to binary finite-precision representations of real number
Embedded Zerotree Wavelet
22 - 3

Overview (2-2) - Embedded Zerotree Wavelet (EZW)

2 Properties
Producing a fully embedded bit stream Providing competitive compression performance

4 Features
Discrete wavelet transform Zerotree coding of wavelet coefficients Successive-approximation quantization (SAQ) Adaptive arithmetic coding

2 Advantages
Precise rate control No training of any kind required
Embedded Zerotree Wavelet
22 - 4

Agenda
Overview Discrete Wavelet Transform Zerotree Coding of Wavelet Coefficients Successive-Approximation Quantization (SAQ) Adaptive Arithmetic Coding Relationship to Other Coding Algorithms A Simple Example Experimental Results Conclusion References Q&A
22 - 5

Embedded Zerotree Wavelet

Discrete Wavelet Transform (2-1)

Identical to a hierarchical subband system
Subbands are logarithmically spaced in frequency Subbands arise from separable application of filters
LL2
HL2

LL1

HL1

HL1

LH2 HH2

LH1

HH1

LH1

HH1

First stage

Second stage

Embedded Zerotree Wavelet

22 - 6

Discrete Wavelet Transform (2-2)

Wavelet decomposition (filters used based on 9-tap symmetric quadrature mirror filters (QMF))
wy

Df1

Df1
wx
In the frequency domain
Embedded Zerotree Wavelet

D32jf

Image wavelet representations

22 - 7

Zerotree Coding (3-1)

A typical low-bit rate image coder
Large bit budget spent on encoding the significance map

Binary decision as to: whether a coefficient has a zero or nonzero quantized value

True cost of encoding the actual symbols: Total Cost = Cost of Significance Map + Cost of Nonzero Values

Embedded Zerotree Wavelet

22 - 8

Zerotree Coding (3-2)

What is zerotree?
A new data structure Parent: IF: To improve the compression of significance maps of The coefficient at the scale coarse scale. wavelet coefficients A wavelet coefficient at a coarse is insignificant with Scanning rule:

Child: respect to a given threshold T, a coefficient xA is said to be insignificant with respect wavelet What is insignificant? No child node is scanned before its parent. coefficient x is All coefficients corresponding to the same spatial location at THEN: to a given threshold T if : Zerotree is element based onof an true hypothesis IF An aempirically zerotree for threshold T is the next finer scale of similar orientation. All itself wavelet coefficients of |x| the same orientation in the same < T are and all of its descendents insignificant with IF Parent-child dependencies (descendants & ancestors) spatial at finerrespect scales are to be zerotree insignificant to Tlikely . It is notlocation the descendant of a previously found root Scanning of the coefficients with respect toTT. for threshold . An element of a zerotree for threshold T Scanning order of the subbands Parent-child dependencies of subbands A zerotree root
Embedded Zerotree Wavelet
22 - 9

Zerotree Coding (3-3)

Encoding

Embedded Zerotree Wavelet

22 - 10

SAQ (3-1)
Successive-Approximation Quantization (SAQ)
Sequentially applies a sequence of thresholds T0,,TN-1 to determine significance

Thresholds
Chose so that Ti = Ti-1 /2 T0 is chosen so that |xj| < 2T0 for all coefficients xj

Two separate lists of wavelet coefficients

Dominant list Subordinate list Dominant list contains: Subordinate list of contains: The coordinates those coefficients that have not yet been found to The magnitudes of those that have been found be significant in the samecoefficients relative order as the initial scan. to be significant.
Embedded Zerotree Wavelet
22 - 11

SAQ (3-2)
Dominant pass a dominant During Subordinate pass pass: with coordinates on the dominant list are compared coefficients Encoding process During a subordinate pass:
to the threshold Ti to determine their significance, and if all coefficients on the subordinate list are scanned and the significant, their sign. SAQ encoding process: specifications of the magnitudes available to the decoder are FOR I = T0 TO TN-1 refined to an additional bit of precision. Dominant Pass; Subordinate Pass (generating string of symbols) ; String of symbols is entropy encoded; Sorting (subordinate list in decreasing magnitude); IF (Target stopping condition = TRUE) break; NEXT;

Embedded Zerotree Wavelet

22 - 12

SAQ (3-3)
Decoding
Each decode symbol, during both passes, refines and reduces the width of the uncertainty interval in which the true value of the coefficient ( or coefficients, in the case of a zerotree root)

Reconstruction value
Can be anywhere in that uncertainty interval Practically, use the center of the uncertainty interval

Good feature
Terminating the decoding of an embedded bit stream at a specific point in the bit stream produces exactly the same image that would have resulted had that point been the initial target rate
22 - 13

Embedded Zerotree Wavelet

Adaptive Arithmetic Coding

Based on [3], encoder is separate from the model
which is basically a histogram

During the dominant passes

Choose one of four histograms depending on
Whether the previous coefficient in the scan is known to be significant Whether the parent is known to be significant

During the subordinate passes

A single histogram is used

Embedded Zerotree Wavelet

22 - 14

Relationship to Other Coding Algorithms

Relationship to Bit Plane Encoding (more general & complex) (all thresholds powers of uncertainty two and all coefficients are a) Reduce the width are of the largest interval in all coefficeints integers ) precision further b) Increase the
c) Attempt to predict insignificance from low frequency to high Most-significant binary digit (MSBD)

Item 1) First b) a) including First a) second b) Dominant bits (digits to second the left and the MSBD) 2) No encoded c) c) Measured and during the dominant pass 3) Training needed No training needed Subordinate bits (digits to the right of the MSBD)
Measured and encoded during the subordinate pass

Its sign and bit position are measured and encoded during the dominant pass PPC EZW

Relationship to Priority-Position Coding (PPC)

22 - 15

Embedded Zerotree Wavelet

Agenda
Overview Discrete Wavelet Transform Zerotree Coding of Wavelet Coefficients Successive-Approximation Quantization (SAQ) Adaptive Arithmetic Coding Relationship to Other Coding Algorithms A Simple Example Experimental Results Conclusion References Q&A
22 - 16

Embedded Zerotree Wavelet

A Simple Example (2-1)

Only string of symbols shown (No adaptive arithmetic coding) Simple 3-scale wavelet transform of an 8 X 8 image T0 = 32 (largest coefficient is 63)
63 -34 49 10 7 13 -12 7 -31 23 14 -13 3 4 6 -1
15 14 3 -12 5 -7 -9 -7 -14 8 4 -2 3 3 9 2

-5 9 -1 47 4 6 -2 2 3 0 -3 2 3 -2 0 4
2 -3
6 4

3 0

6 4
22 - 17

5 11 5
Embedded Zerotree Wavelet

3 -4

Example

A Simple Example (2-2)

First dominant pass First subordinate pass
63 -34 49 10 7 13 -12 7 -31 23 14 -13 3 4 6 -1 15 14 3 -12 5 -7 -9 -7 -14 8 4 -2 3 3 9 2

-5 9 -1 47 4 6 -2 2 3 0 -3 2 3 -2 0 4 2 -3
6 4

3 0

6 4

5 11 5

3 -4

Example
Embedded Zerotree Wavelet

Magnitudes are partitioned into the uncertainty intervals [32, 48) and [48, 64), with symbols 0 and 1.
22 - 18

Experimental Results
12-byte header
Number of wavelet scales For image of Barbara: Dimensions of the image For number of significant coefficients retained at the same low bit rate: in the arithmetic coder Item Maximum histogram JPEG count for the models EZW Same file size meanPSNR lower Looks better Image & initial threshold

Item Other EZW Same PSNR to standard Looks better File size smaller Applied b/w 8 bpp. test images Number retained Less More encoding with the JPEG positions of the significant coefficients.) Compared Compared with other wavelet transform coding
Embedded Zerotree Wavelet

512 X 512 Lena image ( PSNR : Peak-Signal-to-Noise-Rate ) 512 X 512zerotree Barbaracoding image provides a much better way of (Reason: The

22 - 19

Conclusion
2 Properties
Producing a fully embedded bit stream Providing competitive compression performance

4 Features
Discrete wavelet transform Zerotree coding of wavelet coefficients Successive-approximation quantization (SAQ) Adaptive arithmetic coding

2 Advantages
Precise rate control No training of any kind required
Embedded Zerotree Wavelet
22 - 20

References
1. E. H. Adelson, E. Simoncelli, and R. Hingorani, Orthogonal pyramid transforms for image coding, Proc. SPIE, vol.845, Cambridge, MA, Oct. 1987, pp. 50-58 2. S. Mallat, A theory for multiresolution signal decomposition: The wavelet representation, IEEE Trans. Pattern Anal. Mach. Intell., vol. 11, pp. 674-693, July 1989 3. I. H. Witten, R. Neal, and J. G. Cleary, Arithmetic coding for data compression, Comm. ACM, vol. 30, pp. 520-540, June 1987 4. J. Shapiro, Embedded image coding using zerotrees of wavelet coefficients, IEEE Trans. Signal Processing., vol. 41, pp. 3445-3462, Dec. 1993

Embedded Zerotree Wavelet

22 - 21

Embedded Zerotree Wavelet

22 - 22