AU8709798A - A method for digital data compression - Google Patents

Description

ar~-xlmT a41FI~1PI S F Ref: 433354

AUSTRALIA

PATENTS ACT 1990 COMPLETE SPECIFCATION FOR A STANDARD PATENT

ORIGINAL

I -s Name and Address of Applicant: Canon Kabushikl Kaisha 30-2, Shimomaruko 3-chome Ohta-ku Tokyo 146

JAPAN

Actual Inventor(s): Address for Service: Invention Title: Canon Information Systems Research Australia Pty Ltd 1 Thomas Holt Drive North Ryde New South Hales 2113

AUSTRALIA

James Philip Andrew Spruson Ferguson, Patent Attorneys Level 33 St Martins Tower, 31 Market Street Sydney, New South Wales, 2000, Australia A Method for Digital Data Compression ASSOCIATED PROVISIONAL APPLICATION DETAILS [31] Application No(s) [33] Country P09513 AU PP0776 AU [32] Application Date 29 September 1997 8 December 1997 ;1 .1 ;i 8; ii: i,.

rj :i; The following statement is a full description of this invention, including the best method of performing it known to me/us:- -1- A Method for Digital Data Compression Field of Invention The present invention relates to the field of data compression with particular application to digital image compression. More particularly, the present invention discloses a digital image compression method using spatial magnitude context entropy coding of discrete wavelet transform coefficients.

Background of Invention The field of digital data compression and in particular digital image compression has attracted great interest for some time.

In the field of digital image compression, many different techniques have been utilised. In particular, one popular technique is the JPEG standard which utilises the discrete cosine transform to transform standard size blocks of an image into corresponding cosine components. In this respect, the higher frequency cosine components are heavily quantised so as to assist in obtaining substantial compression factors. The heavy quantisation is an example of a "lossy" technique of image compression. The JPEG standard also provides for the subsequent lossless compression 20 of the transformed coefficients.

Recently, the field of wavelet transforms has gained great attention as an alternative form of data compression. The wavelet transform has been found to be highly suitable in representing data having discontinuities such as sharp edges. Such discontinuities are often present in image data or the like.

Although the preferred embodiments of the prtsent invention will be described with reference to the compression of image data, it will be readily evident that the preferred embodiment is not limited thereto. For examples of the many different applications of Wavelet analysis to signals, reference is made to a survey article entitled "Wavelet Analysis" by Bruce et. al. appearing in IEEE Spectrum, October 1996 page 26 35. For a discussion of the different applications of wavelets in computer graphics, reference is made to "Wavelets for Computer Graphics", I. Stollinitz et. al. published 1996 by Morgan Kaufmann Publishers, Inc.

Aspects of Invention It is an object of the present invention to ameliorate one or more disadvantages a of the prior art.

According to a first aspect of the present invention, there is provided a method of compressing data, the method including the steps of: V Lw

I

-2applying a transform to the data to produce transformed data having a series of parts; entropy encoding the magnitude of the transformed data of at least one of said parts based on the magnitude of the surrounding transformed data; and separately encoding the value of said transformed data.

According to a second aspect of the present invention, there is provided a method of decompressing data, the method including the steps of entropy decoding a portion of the data to generate the magnitudes of transform data based on the magnitudes of surrounding, previously entropy decoded, transformed data; and separately decoding another portion of the data to generate the values of said transformed data; and applying an inverse transform to the transform data to generate decompressed data.

According to a third aspect of the present invention, there is provided an apparatus for compressing data, the apparatus including: transform means for applying a transform to the data to produce transformed data having a series of parts; entropy encoding means for entropy encoding the magnitude of the 2C transformed data of at least one of said parts based on the magnitude of the surrounding Stransformed data; and encoding means for separately encoding the value of said transformed data.

According to a fourth aspect of the present invention, there is provided an apparatus for decompressing data, the apparatus including; entropy decoder means for entropy decoding a portion of the data to generate the magnitudes of transform data based on the magnitudes of surrounding, previously entropy decoded, transformed data; and i~ decoder means for separately decoding another portion of the data to generate the values of said transformed data; and inverse transform means for applying an inverse transform to the transform data to generate decompressed data.

According to a fifth aspect of the present invention, there is provided a computer program product including a computer readable medium having recorded -thereon a computer program for compressing data, the computer program product including: transform means for applying a transform to the data to produce ;ransformed data having a series of parts; -3entropy encoding means for entropy encoding the magnitude of the transformed data of at least one of said parts based on the magnitude of the surrounding transformed data; and encoding means for separately encoding the value of said transformed data.

According to a sixth aspect of the present invention, there is provided a computer program product including a computer readable medium having recorded thereon a computer program for decompressing data, the computer program product including: entropy decoder means for entropy decoding a portion of the data to generate 1o the magnitudes of transform data based on the magnitudes of surrounding, previously entropy decoded, transformed data; and decoder means for separately decoding another portion of the data to generate the values of said transformed data; and inverse transform means for applying an inverse transform to the transform data to generate decompressed data.

According to a seventh aspect of the present invention, there is provided a method of compressing data, the method including the steps of: applying a transform to the data to produce a plurality of transform coefficients, wherein each transform coefficient is expressible by a code representation including a plurality of symbols; entropy encoding one of said symbols, not previously entropy coded, of a current transform coefficient based on a context of surrounding symbols; repeating step a predetermined number of times for the current transform coefficient; and 25 processing another transform coefficient in accordance with steps and According to a eighth aspect of the present invention, there is provided a Ii method of compressing data including the steps of: a) applying a transform to the data to produce a plurality of transform coefficients, wherein each transform coefficient is expressible by a binary code representation having a plurality of bits; b) entropy encoding one of said bits, not previously entropy coded, of a current transform coefficient based on a context of surrounding bits; c) repeating step b) a predetermined number of times the current transform coefficient; and d) processing another transform coefficient in accordance with steps b) and 1 i i -b I -4- According to a ninth aspect of the present invention, there is provided a method of decompressing data, the method including the steps of: entropying decoding said data to generate a symbol of a current transform coefficient based on a context of surrounding symbols; repeating step a predetermined number of times for the current transform; generating another transform coefficient in accordance with steps and and applying an inverse transform to the transform coefficients to produce data.

According to a tenth aspect of the present invention, there is provided a method of decompressing data, the method including the steps of: entropying decoding said data to generate a bit of a current transform coefficient based on a context of surrounding bits; repeating step a predetermined number of times for the current transform; generating another transform coefficient in accordance with steps and b applying an inverse transform to the transform coefficients to produce data.

S. According to an eleventh aspect of the present invention, there is provided an apparatus for compressing data, the apparatus including: transform means for applying a transform to the data to produce a plurality of transform coefficients, wherein each transform coefficient is expressible by a code representation including a plurality of symbols; entropy encoder means for entropy encoding one of said symbols, not previously entropy coded, of a current transform coefficient based on a context of surrounding symbols; repetition means for repeating the operations of the entropy encoder means a predetermined number of times for the current transform coefficient; and processor means for processing another transform coefficient in accordance with the operations of the entropy encoder means and repetition means.

According to a twelfth aspect of the present invention, there is provided an i .apparatus for compressing data including: ""transform means for applying a transform to the data to produce a plurality of transform coefficients, wherein each transform coefficient is expressible by a binary code representation having a plurality of bits; entropy encoder means for entropy encoding one of said bits, not previously entropy coded, of a current transform coefficient based on a context of surrounding bits; repetition means for repeating the operation of the entropy encoder a predetermined number of times for the current transform coefficient; and processor means for processing another transform coefficient in accordance with the operations of the entropy encoder means and repetition means.

s According to a thirteenth aspect of the present invention, there is provided an apparatus for decompressing data, the apparatus including: entropy decoder means for entropying decoding said data to generate a symbol of a current transform coefficient based on a context of surrounding symbols; repetition means for repeating the operation of the entropy decoder means a predetermined number of times for the current transform coefficient; generation means for generating another transform coefficient in accordance with the operations of the entropy decoder means and the repetition means; and inverse transform means for applying an inverse transform to the transform coefficients to produce data.

According to a fourteenth aspect of the present invention, there is provided an apparatus for decompressing data, the apparatus including: entropy decoder means for entropying decoding said data to generate a bit of a current transform coefficient based on a context of surrounding bits; repetition means for repeating the operation of the entropy decoder means a predetermined number of times for the current transform coefficient; generation means for generating another transform coefficient in accordance with the operations of the entropy decoder means and the repetition means; and inverse transform means for applying an inverse transform to the transform coefficients to produce data.

According to a fifteenth aspect of the present invention, there is provided a computer program product including a computer readable medium having recorded thereon a computer program for compressing data, the computer program product including: transform means for applying a transform to the data to produce a plurality of transform coefficients, wherein each transform coefficient is expressible by a code representation including a plurality of symbols; entropy encoder means for entropy encoding one of said symbols, not previously entropy coded, of a current transform coefficient based on a context of surrounding symbols; and repetition means for repeating the operations of the entropy encoder means a predetermined number of times for the current transform coefficient; and processor means for processing another transform coefficient in accordance with the operations of the entropy encoder means and repetition means.

-6- According to a sixteenth aspect of the present invention, there is provided a computer program product including a computer readable medium having recorded thereon a computer program for compressing data, the computer program product including: transform means for applying a transform to the data to produce a plurality of transform coefficients, wherein each transform coefficient is expressible by a binary code representation having a plurality of bits; entropy encoder means for entropy encoding one of said bits, not previously entropy coded, of a current transform coefficient based on a context of surrounding Io bits; repetition means for repeating the operation of the entropy encoder a predetermined number of times the current transform coefficient; and processor means for processing another transform coefficient in accordance with the operations of the entropy encoder means and repetition means.

According to a seventeenth aspect of the present invention, there is provided a computer program product including a computer readable medium having recorded thereon a computer program for decompressing data, the computer program product including: Sentropy decoder means for entropying decoding said data to generate a symbol 20 of a current transform coefficient based on a context of surrounding symbols; repetition means for repeating the operation of the entropy decoder means a predetermined number of times for the current transform coefficient; generation means for generating another transform coefficient in accordance with the operations of the entropy decoder means and the repetition means; and inverse transform means for applying an inverse transform to the transform coefficients to produce data.

According to an eighteenth aspect of the present invention, there is provided a computer program product including a computer readable medium having recorded thereon a computer program for decompressing data, the computer program product including: .entropy decoder means for entropying decoding said data to generate a bit of a current transform coefficient based on a context of surrounding bits; repetition means for repeating the operation of the entropy decoder means a predetermined number of times for the current transform coefficient; a 35 generation means for generating another transform coefficient in accordance *with the operations of the entropy decoder means and the repetition means; and inverse transform means for applying an inverse transform to the transform coefficients to produce data.

sL -7- Brief Description of the Drawings Embodiments of the invention are described, by way of example only, with [I reference to the drawings, in which: S Figs. 1-3 illustrate the process of wavelet transforming image data; Fig. 4 illustrates the steps involved in the encoder of the first preferred embodiment; Fig. 5 illustrates the steps in the decoder as constructed in accordance with the first preferred embodiment; Fig. 6 illustrates a data structure utilised by the first preferred embodiment; Fig. 7 illustrates the process of utilising a surrounding context for a current coefficient in accordance with the first preferred embodiment; Fig. 8 illustrates the process of histogram calculation for each subband; Fig. 9 illustrates the steps involved in the encoder of the second preferred embodiment; Fig. 10 illustrates the steps in the decoder as constructed in accordance with the second preferred embodiment; Fig. 11 illustrates a data structure utilised by the second preferred embodiment; Fig. 12 illustrates a window for determining a context for a current coefficient in accordance with the second preferred embodiment; Fig. 13 illustrates the process of utilising surrounding context for a current coefficient at bitplane n; Fig. 14 illustrates the process of block based encoding in accordance with a further embodiment of the invention; and Fig. 15 is a block diagram of a general purpose computer; 3.

e

L

Detailed Description The preferred embodiments proceed initially by means of a wavelet transform of image data. A description of the wavelet transform process is given in many standard texts and in particular the aforementioned book by Stollnitz et. al. Before proceeding with a description of the preferred embodiments, a brief overview of the wavelet process will now be described with reference to the accompanying drawings.

Referring initially to Fig. i, an original image 1 is transformed utilising a Discrete Wavelet Transform (DWT) into four subimages 3-6. The subimages or subbands are normally dencied LL1, HL1, LH1 and HH1. The one suffix on the subband names indicates level 1. The LL1 subband is a low pass decimated version of the original image.

S-8- The wavelet transform utilised can vary and can include, for example, Haar basis functions, Daubechies basis functions etc. The LL1 subband is then in turn utilised and a second Discrete Wavelet Transform is applied as shown in Fig. 2 giving subbands LL2 HL2 LH2 HH2 This process is continued for example as illustrated in Fig. 3 wherein the LL4 subband is illustrated, the LL4 band decomposition process being referred to as an octave band filter bank with the LL4 subband being referred to as the DC subband. Obviously, further levels of decomposition can be provided depending on the size of the input image.

Each single level DWT can in turn be inverted to obtain the original image.

o1 Thus a J-level DWT can be inverted as a series of J-single level inverse DWT's.

To code an image hierarchically the DC subband is coded first. Then, the remaining subbands are coded in order of decreasing level. That is for a 4 level DWT, the subbands at level 4 are coded after the DC subband (LL4). That is the HL4, LH4 and HH4 subbands. The subbands at level 3 (HL3, LH3, and HH3) are then coded, 1i followed by those at level 2 (HL2, LH2 and HH2) and then level 1 (HL1, LH1 and

HHI).

With standard images, the encoded subbands normally contain the "detail" information in an image. Hence, they often include a sparse array of values and substantial compression can be achieved by quantisation of the subbands and efficient 20 encoding of their sparse matrix form.

1.0 First Preferred Embodiment S" In the first preferred embodiment, an effective compression of the subbands is Sprovided through the utilisation of the correlation between the energy of adjacent S 25 coefficients in the DWT subband- The encoding proceeds via two stages. The first stage encodes the leading zeros, or magnitude category of each DWT coefficient, based on the context of the number of surrounding coefficients that are not quantised to zero. By using a limited window this number takes on a relatively small range of values, and hence there are a limited number of contexts. For efficient entropy coding such a small number of contexts is desired. After the number of leading zero is transmitted or encoded the remaining bits for coefficients that are not quantised to zero are transmitted or encoded as is. Entropy coding could also be used at this stage. However, experiments suggest that this may not substantially increase the compression, while introducing more complexity.

An overview of the coding process is illustrated 20 in Fig. 4, while the decoding process is illustrated 30 in Fig. -9- Turning initially to Fig. 4, a digital image is transformed 21 using a Discrete Wavelet Transform into several subband components as previously described. Each subband is preferably coded in a hierarchical order. As illustrated in Fig. 6, each coefficient in a subband is quantised to an integer value 25 having a predetermined number of bits L, and conceptually represented in a binary format with a sign bit s.

This integer is represented using the number of leading zeros Z from a predetermined maximum bit number and the remaining bits r with a sign bit. This number of leading zeros Z is entropy coded based on the number of surrounding coefficients within a certain window that are not quantised to zero. For non zero integer coefficients the remaining bits r and sign bit s are coded as is.

As illustrated in Fig. 5, at the decoder the operation of the encoder is reversed (in as much as this is possible with quantisation). The quantised coefficients are inverse quantised 31. Finally an inverse Discrete Wavelet Transform is performed 32 on the resulting subbands to give the output image.

is As noted previously, in the encoding process the discrete wavelet transform coefficients are quantised 22 to integer values. Let c represent a coefficient value and d its quantised values. Then the quantisation is performed as, S" where q is a predetermined quantisation factor and fix is defined by, f 20 r ixx0 were L is the round down to nearest integer operator and F is the round I up to nearest integer operator. At the encoder each coefficient in a subband is quantised to an integer value using this equation.

The inverse quantisation is given by, c= q x d ign(d) x where, I" -lad<0 sign(d) 0 d= 3, At the decoder each coefficient is inverse quantised using this inverse quantisation equation. The quantisation factor q can vary from subband to subband, or 30 it can be fixed for the whole image. It can be coded in the header of the compressed image.

M m

I

1.1 Coefficient Coding and Decoding As shown in Fig. 6, each quantised coefficient is an integer value represented in a binary format with a sign bit. For the purpose of the description of the preferred embodiment, it is assumed with 15 bits and an extra sign bit (ie. L= Thus sign(d) x bb -bo where bn is binary bit n. If the coefficient is non zero and the most significant bit number 26 is m, then: bb

I

and the number of leading zeros Z is, o1 Z= 14-m, If the coefficient is zero we set Z= 15. The coefficient d is coded in two parts.

First Z is entropy coded based on the context of the number of surrounding coefficients that are non-zero. A concise definition of surrounding coefficients is given below.

Then for non-zero coefficients bits and the sign bit are coded into the bit is stream.

1.2 Surrounding Coefficient Context Turning to Fig. 7. a subband eg. 35 is coded in raster scan order from top to bottom and left to right. If the current coefficient to be coded is marked with a cross zo 36, the surrounding coefficients are considered to be the four surrounding coefficients indicated by the four empty squares 37-40. The surrounding coefficients are selected by a window with a shape as indicated in Fig. 7. If the cross in the window is aligned with the current coefficient the surrounding coefficients 37-30 are defined to be the coefficients that fall within the window.

The window illustrated follows a raster scan order. Hence, when the current coefficient is being decoded the surrounding coefficients have already been decoded, and thus the decoder knows whether or not they are non zero. The context for the current pixel is determined by the number of surrounding coefficients 37-70 that are non-zero. In this case there are 5 contexts corresponding to 0, 1, 2, 3 or 4 surrounding pixels that are non-zero. For the coefficients in the first row or column a modified window is used that includes only coefficients in the current subband.

Obviously different windows can be used. Ideally, the surrounding coefficients must come before the current coefficient in the raster scan order. This is so the decoder knows their value before decoding the current pixel. The four coefficient window of Fig. 7 has been selected for a good compromise between complexity, which grows with the number of contexts, and compression efficiency, which increases with the number of contexts, at least up to a certai point_

L

1* v^' is nr oo~urnw -11- 1.3 Context Entropy Coding As noted previously, the number of leading zeros Z of each coefficient is coded with a context based entropy coder. Preferably, this is a standard arithmetic coder. Arithmetic coding is described in Witten et. al., "Arithmetic coding for data compression", Communications of the ACM, Volume 30, No. 6, June 1987.

Preferably fixed histograms are utilised. The histograms are then coded into the compressed image header and are used by the decoder.

In the preferable implementation the arithmetic coder is initialised with different histograms for each subband. For a given subband a histogram with 16 bins is lo generated for each context as illustrated in Fig. 8. For a given context the ith bin of the histogram is the count of the number of coefficients with i leading zeros and whose context of surrounding coefficients is the given context. For example, for a context of 2 surrounding non-zero coefficients, bin 7 of the histogram is the count of the number of coefficients with 7 leading zeros and who have 2 surrounding non-zero pixels.

These histograms are made prior to the coding process. As an alternative an adaptive arithmetic coder could be utilised.

At the decoder these histograms are used to decode the leading zero information for the subband coefficients, which is contained in the coded bit stream.

For the non-zero coefficients, the remaining bits and sign bit are read from the coded bit stream and hence the quantised coefficients can be constructed.

Second Preferred Embodiment In the second preferred embodiment, an effective compression of the subbands is also provided through the utilisation of relationship between the energy of adjacent coefficients in the DWT subband.

The encoding proceeds via a depth first approach. That is, each bit of a current coefficient is encoded, based on the context of bits, of previously coded surrounding coefficients, on a current bit-plane (bit-plane number n) and on whether or not a most significant bit (msb)of the current coefficient has been coded (ie. the msb number is greater than bit-plane By using a limited window a number contexts used in the encoding takes on a relatively small range of values, and hence there are a limited number of contexts. For efficient entropy coding such a small number of contexts is desired. After substantially all the bits of the current coefficient are encoded, a sign (positive or negative of the current coefficient, if the current coefficient is not zero, is transmitted or encoded as is. Optionally entropy coding of a sign of the coefficient could also be used at this stage.

An overview of the coding process is illustrated 120 in Fig. 9, while the decoding process is illustrated 130 in Fig. 10, in the form of structure diagrams.

i.

i r, -r

ID

i i~ _i- -12- Typically structure diagrams a read with a left operation item begin performed before a right operation item.

Turning initially to Fig. 9, a digital image is transformed 121 using a Discrete Wavelet Transform into several subband components as previously described. Each b subband is preferably coded in a hierarchical order. An encoding coding of the transform coefficients is performed 143 by a binary arithmetic coder. Each coefficient 144 in each subband 145 is iteratively processed by a quantisation 122 then an entropy coding by a binary arithmetic coder 143. As illustrated in Fig. 11, each coefficient in a subband is quantised to an integer value 125 having a predetermined number of bits L, and conceptually represented in a binary format with a sign bit s. Each bit 161 is preferably coded in order from a most significant bit-plane to a least significant bitplane. A bit 161 of the integer value 125 is entropy coded using a context formed from a consideration of surrounding bits, and whether or not a msb of the integer value 125 has been entropy coded.

As illustrated in Fig. 10, at the decoder the operation of the encoder is reversed (in as much as this is possible with quantisation). A predetermined portion of an encoded bit-stream (the result of the coder) is decoded 150 by a binary arithmetic i decoder to provide a quantised coefficient. The quantised coefficient is inverse quantised 131. This process is looped through 151,152 in an, iterative manner, to produce each coefficient of each subband. Finally an inverse Discrete Wavelet Transform is performed 132 on the resulting subbands to give the output image. This process need not be iterative. For example, the entire bitstream can be encoded by the binary arithmetic decoder 133 to provide substantially all the quantised coefficients and then these coefficients can be inverse quantised 31 before applying the inverse Discrete Wavelet Transform 132.

As noted previously, in the encoding process the discrete wavelet transform coefficients are quantised 122 to integer values. Let c represent a coefficient value and d its quantised values. Then the quantisation is performed as, d=fix\where q is a predetermined quantisation factor and fix is defined by, lx 0 flx(X)= {1 Z 0 were L is the round down to nearest integer operator and [1 is the round up to nearest integer operator. At the encoder each coefficient in a subband is quantised to an integer value using this equation.

The inverse quantisation is given by,

JI

F^ -13c q x d sign(d) x 2 where, II sign(d)= 0 d S1 d>O At the decoder each coefficient is inverse quantised using this inverse quantisation equation. The quantisation factor q can vary from subband to subband, or it can be fixed for the whole image. It can be coded in the header of the compressed image.

2.1 Coefficient Coding and Decoding As shown in Fig. 11, each quantised coefficient is an integer value represented in a binary format with a sign bit. For the purpose of the description of the preferred embodiment, it is assumed with 15 bits and an extra sign bit (ie. Thus d sign(d) x b 4 bo where bn is binary bit n.

The coefficient d is coded by entropy coding the bits b 14 b 0 in order, and a sign bit i for non-zero coefficients. Bit bn of a coefficient is coded with a binary entropy coder i (eg. binary arithmetic coder) based on the context determined by a bit pattern formed from: bit n of each of the surrounding coefficients; whether the most significant bit of a current coefficient has been coded (ie whether the msb number is greater than and on whether the msb of any of the surrounding coefficients have been coded. A description of surrounding coefficients and context thus formed is given below. A sign of each coefficient not quantised to zero can be ceded as is or entropy coded based on surrounding coefficient signs.

At the decoder the quantised coefficient d is reconstructed by simply entropy 8 (eg. binary arithmetic) decoding bits b 14 b 0 and a sign bit for non-zero coefficients.

2.2 Surrounding Coefficient Context Turning to Fig. 12, a subband eg. 135 is coded in raster scan order from top to S 30o bottom and left to right. If the current coefficient to be coded is marked with a cross 136, the surrounding coefficients are considered to be the four surrounding coefficients indicated by the four empty squares 137-140. Preferably the surrounding coefficients are selected by a window with a shape as indicated in Fig. 12. If the cross in the window is aligned with the current coefficient the surrounding coefficients 137-140 are defined to be the coefficients that fall within the window. !i n -14- The window illustrated follows a raster scan order. Hence, when the current coefficient is being decoded the surrounding coefficients have already been decoded, and thus the decoder knows whether or not they are non zero.

Referring to Fig. 13, there is shown a single bit-plane 150, bit-plane n, of the DWT coefficients. The window, described with reference to Fig. 12, includes four bits at bit-plane n labelled c2 c5 of surrounding coefficients 142-145. Preferably additional flags other than the surrounding bits c2 c 5 are used to determine a context. The additional flags can include a flag, c, that indicates whether or not a current coefficient 141 has a msb that has been previously entropy coded, and/or a flag, ci, indicating whether or not any the surrounding coefficients have a msb which has been previously entropy coded. That is, if encoding the coefficient is performed from a highest value bit-plane to a lowest value bit-plane than c o represent whether or not there is a set bit of a current coefficient in bit-planes above the current bit-plane, and c 1 represent whether or not there is a set bit on any of the surrounding coefficient 142-145 in bit-planes above or in the current bit-plane.

A context for bit n of the current coefficient 141 is determined, in the second preferred embodiment, by 6 bit binary number co, c 1 c 5 Bit cg of the context is set (ie has value 1) if the most significant bit (msb) of the current coefficient has Vg ^already been coded. That is the msb of the current coefficient is in bit plane n +1 or 20 greater. Bit c 1 is set if any one of the four surrounding coefficients 142-145 has a msb in bit plane n or greater. Finally bits c2, c3, c 4 and c5 are determined by a bit pattern of the four surrounding coefficients 142-145. In this case, for example, there are 26 64 contexts, since there are 64 different permutations of binary bits co, cl, The context formation described above is causal in that the context can be formed from previously coded information. In this manner the decoder can form the same context as the encoder. For the window in Fig. 13, following the raster scan order 146, when bit n of the current coefficient is being decoded the surrounding coefficients 142-145 have already been decoded, and thus the decoder knows bit n for each of these coefficients, 142-145, and whether or not the msb of these coefficients is in a bit-plane greater than or equal to bit-plane n. The decoder also knows whether the msb of the current coefficient is in a bit-plane greater than or equal to bit-plane n+l.

Thus the decoder can form the same context as used at the encoder.

Optionally different windows and different information about previously coded bits in coefficients can be used to form a variety of different sets of contexts without departing from the scope and spirit of the invention. For example another window configuration can be used to determine a context. Typically information used to determine a context is causal. That is, surrounding coefficients (or bits) used in the I determination of the context are to be processed before a current coefficient.

prcse a' curn coefficient Preferably in a raster scan (or coding) order. A small local window is also desired because it minimises the local memory buffering requirements. In addition, a small number of contexts is preferred to minimise the amount of memory required for the entropy coding, and to prevent context dilution. The four coefficient window, and s whether or not the msb has been coded, is used, herein, as a good compromise between complexity, which grows with the number of contexts, and compression efficiency, which increases with the number of contexts, at least up to a predetermined number of such contexts.

Typically, once the most significant bit of a coefficient has been coded, a different context entropy coder could be used: one that is based more on the current coefficient to be coded than on the surrounding coefficients. Optionally, it is possible to simple code these bits without entropy coding.

2.3 Context Entropy Coding 15 As noted previously, each bit of a current coefficient is coded with a context based entropy coder. Preferably, this is a standard arithmetic coder. Arithmetic coding is described in Witten et. al., "Arithmetic coding for data compression", Communications of the ACM, Volume 30, No. 6, June 1987.

Typically Arithmetic coding relies on assigning a probability of occurrence of "20 a symbol, in a plurality of symbols to be encoded. Hence one option of the preferred S embodiment is to assign a predetermined probability to each of the 64 different contexts, preferably so that each context has a probability value indicative of a likelihood that a current symbol has a resulting value. The contexts and their S" "corresponding probabilities form a look up table that is duplicated for a decoder, so the decoder can mimic the encoder. Naturally, a fixed probability approach works well on stationary images but in reality not all images are stationary and while a performance of the encoderldecoder is adequate, using a fixed probability approach, both for stationary i and non-stationary sources an adaptive arithmetic coder is preferred. For non- Sstationary sources (images) an adaptive arithmetic (de)coder will overcome changes in the probability distribution of each symbol. Another embodiment of the present invention can be described as a variation on the second preferred embodiment in which a combination of a breadth first and a depth first approach is adopted. In the present embodiment an image is, transformed, quantised and divided into a plurality of blocks. Preferably each block including a fixed number of quantised transformed coefficients. A predetermined block of transformed coefficients is processed in a breadth first manner. That is, each i -i coefficient in the block and each symbol of each coefficient is processed in a sequential manner. For example a first symbol of a first coefficient is encoded substantially as "j sg~~ -16described in the preferred embodiment, then a first symbol of a second coefficient is encoded, then a first symbol of a third coefficient is encoded etc. Until all first symbols of the coefficients in the block have been encoded. Next all the second symbols of each coefficient in the block are encoded. This is continued across (breadth s first) each coefficient in the block until substantially all coefficients and their corresponding symbols have been encoded. Another block of coefficients is then processed in a substantially similar manner until substantially all the blocks and hence substantially all coefficients of the transformed image are encoded.

Essentially, the depth first approach upon each transform coefficient described, above, with reference to the second preferred embodiment is adopted on a block by block basis. However, the coefficients within each block are encoded in a breadth first Happroach.

Referring to Fig. 14, there is shown an array 160 of transform coefficients 161, representing a set of quantised transform coefficients 161 of an image. The array i 160 of coefficients 161 are divided into a plurality of blocks and an example of one such block 162 is also shown in Fig. 14. The block 162 is a 2x2 block including four coefficients 163-166 which are typically encoded on a symbol by symbol basis in a zigzag fashion 167. That is, a first symbol of the first coefficient 163 of the block 162 is encoded based on a context determined for the first symbol substantially as described in the second preferred embodiment. Next a first symbol of the second coefficient 164 of the block 162 is encoded. Followed by an encoding of a first symbol of the third coefficient 165 of the block 162 and similarly a first symbol of the fourth coefficient 166 is encoded. Next a second symbol of all the coefficients 163-166, of the block o 162, are encoded in substantially the same manner. This is repeated until substantially all symbols of the coefficient of the block 162 are encoded before another block is -M encoded.

A determination of an ideal block size (ie. the number of coefficients per J, block) may vary from application to application and/or an amount of available high speed cache memory, however a preferred block size is one than exploits both advantages of a depth first approach and a breadth first approach. Advantages of the breadth first approach include having knowledge of previously encoded symbols of an entire block. For example, as noted previously, once all the first symbols of each coefficients of a block is encoded a context of second symbols of each coefficient can benefit from a knowledge of the first symbols encoded across the entire block.

35 Advantages of a depth first approach include on each block is that an array of coefficients can be read on a block by block basis rather than a symbol by symbol basis Srequired in a purely breadth first approach, thus reduce the number of accesses to the array.

"DS "~a-WIeB rS:.BO -17- Typically, the array 160 of coefficients is stored in memory such as RAM or on external memory such as a hard disk drive for example. Access time for such memory read and/or writes is typically long when compared to on processor cache memory. Consequently, increased performance can be had by caching the block 162 in local cache (high speed) memory rather than access coefficients from the array 160 stored in a hard disk drive or lower performance memory. Images are generally large with respect to the number of bytes per file when compared to, for instance, a text file and hence it is not unreasonable to store a library of images on a hard disk drive. The method of the present embodiment therefore reduces the number of accesses to the array 160 stored in external memory than the number of accesses of a purely breadth first approach. Thus improving performance without the need to cache substantially the entire array 160.

A context of a current symbol in the block based coding technique of the present embodiment is determined substantially as hereinbefore described with reference to the preferred embodiment. Thus in addition to caching the block 162, surrounding coefficients 168 are also cached. Naturally, the number of coefficients 168 and which coefficients are to be cached, in addition to the block 162, depends upon a choice of window and/or flag bits for determining a context.

The preferred embodiments also have application to other image formats. For example, full colour images can be encoded via separate colour channels or the usual chrominance compression techniques as utilised in the JPEG standard can be applied so as to produce reduced chrominance data. Further, the preferred embodiments are described with reference to binary symbols (ie 1 or 0) and bitplanes, however the embodiments, ie the depth first approach, can also be implemented with a n-ary representation of coefficients. That is, each coefficient can be represented by a plurality of different symbols ie. N symbols for an n-ary representation). Still further, not all I" symbols of a current coefficient need be encoded before encoding a next coefficient.

For example, in an eight bit representation of coefficients, four bits of each coefficient :i can be encoded on each pass (raster scan). Hence in two passes of memory each coefficient will be encoded.

-Additionally, the principles of the preferred embodiments can be equally extended to other forms of data such as sound data etc. and the preferred embodiments have application wherever wavelet transforms are suitable. Additionally, the preferred embodiments can be applied to other forms of transformed data for example, the discrete cosine transform process in addition to the wavelet packet and cosine packet transform techniques as described in the aforementioned survey article. -18- 3.1 Preferred Embodiment of Apparatus(sl The encoding and/or decoding process are preferably practiced using a conventional general-purpose computer, such as the one shown in Fig. 15, wherein the processes of Fig. 4 to 14 may be implemented as software executing on the computer.

9 In particular, the steps of the encoding and/or decoding methods are effected by instructions in the software that are carried out by the computer. The software may be divided into two separate parts; one part for carrying out the encoding and/or decoding methods; and another part to manage the user interface between the latter and the user.

The software may be stored in a computer readable medium, including the storage devices described below, for example. The software is loaded into the computer from the computer readable medium, and then executed by the computer. A computer readable medium having such software or computer program recorded on it is a computer program product. The use of the computer program product in the computer preferably effects an advantageous apparatus for encoding and/or decoding representations of digital images in accordance with the embodiments of the invention.

The computer system 1500 includes of the computer 1502, a video display 1516, and input devices 1518, 1520. In addition, the computer system 1500 can have any of a number of other output devices including line printers, laser printers, plotters, and other reproduction devices connected to the computer 1502. The computer system 1500 can be connected to one or more other computers via a communication interface 1508c using an appropriate communication channel 1530 such as a modem communications path, a computer network, or the like. The computer network may I: include a local area network (LAN), a wide area network (WAN), an Intranet, and/or the Internet The computer 1502 itself includes of a central processing unit(s) (simply referred to as a processor hereinafter) 1504, a memory 1506 which may include random access memory (RAM) and read-only memory (ROM), input/output (10) interfaces .1508a, 1508b 1508c, a video interface 1510, and one or more storage devices S. generally represented by a block 1512 in Fig. 15. The storage device(s) 1512 can S" 30 include of one or more of the following: a floppy disc, a hard disc drive, a magnetooptical disc drive, CD-ROM, magnetic tape or any other of a number of non-volatile storage devices well known to those skilled in the art. Each of the components 1504 to 1512 is typically connected to one or more of the other devices via a bus 1514 that in turn can include of data, address, and control buses.

The video interface 1510 is connected to the video display 1516 and provides video signals from the computer 1502 for display on the video display 1516. User input to operate the computer 1502 can be provided by one or more input devices 'k' S -19- 1508b. For example, an operator can use the keyboard 1518 and/or a pointing device such as the mouse 1520 to provide input to the computer 1502.

The system 1500 is simply provided for illustrative purposes and other configurations can be employed without departing from the scope and spirit of the invention. Exemplary computers on which the embodiment can be practiced include IBM-PCIATs or compatibles, one of the Macintosh (TM) family of PCs. Sun Sparcstation or the like. The foregoing is merely exemplary of the types of computers with which the embodiments of the invention may be practiced. Typically, the processes of the embodiments, described hereinafter, are resident as software or a program recorded on a hard disk drive (generally depicted as block 1512 in Fig. 15) as the computer readable medium, and read and controlled using the processor 1504.

Intermediate storage of the program and pixel data and any data fetched from the network may be accomplished using the semiconductor memory 1506, possibly in concert with the hard disk drive 1512.

In some instances, the program may be supplied to the user encoded on a CD- ROM or a floppy disk (both generally depicted by block 1512), or alternatively could be read by the user from the network via a modem device connected to the computer, for example. Still further, the software can also be loaded into the computer system 1500 from other computer readable medium including magnetic tape, a ROM or 20 integrated circuit, a magneto-optical disk, a radio or infra-red transmission channel between the computer and another device, a computer readable card such as a PCMCIA card, and the Internet and Intranets including email transmissions and information I recorded on websites and the like. The foregoing is merely exemplary of relevant computer readable mediums. Other computer readable mediums may be practiced 25 without departing from the scope and spirit of the invention.

The method of encoding and/or decoding may alternatively be implemented in dedicated hardware such as one or more integrated circuits performing the functions or sub functions of the encoding and/or decoding. Such dedicated hardware may include graphic processors, digital signal processors, or one or more microprocessors and associated memories.

The foregoing only describes a small number of embodiments of the present invention, however, modifications and/or changes can be made thereto by a person skilled in the art without departing from the scope and spirit of the invention. The present embodiments are, therefore, to be considered in all respects to be illustrative 1| and not restrictive.

I

J

Claims (79)

1. A method of compressing data, the method including the steps of: applying a transform to the data to produce transformed data having a series of parts; entropy encoding the magnitude of the transformed data of at least one of said parts based on the magnitude of the surrounding transformed data; and separately encoding the value of said transformed data.

2. A method as claimed in claim 1, wherein said entropy encoding utilizes the number of non-zero coefficients surrounding a spatial location of a corresponding transformed data value.

3. A method as claimed in claim 2, wherein the surrounding values utilized are previously entropy encoded values adjacent to a current spatial location of the corresponding transformed data value.

4. A method as claimed in claim 1, wherein said transforming step further includes quantizing transformed portions of said data to integer values. A method as claimed in claim 4, wherein said integer values include a sign bit 20 and a predetermined number of coefficient bits.

6. A method as claimed in 1, wherein said transform includes wavelet -t transforming the data. 25 7. A method as claimed in claim 6, wherein said parts include each of the sub- band components of the wavelet transform which are separately entropy encoded.

8. A method as claimed in claims 6 or 7, wherein the lowest frequency sub-band component is separately encoded.

9. A method as claimed in claim 1, wherein said data includes image data S: describing an image- A method as claimed in claim 1, wherein said magnitude encoding includes 35 encoding the number of leading zeros in transformed data values- J 1 11. A method of decompressing data, the method including the steps of S'" -21 entropy decoding a portion of the data to generate the magnitudes of transform data based on the magnitudes of surrounding, previously entropy decoded, transformed data; and separately decoding another portion of the data to generate the values of said transformed data; and applying an inverse transform to the transform data to generate decompressed data.

12. A method as claimed in claim 11, wherein said entropy decoding utilizes the o1 number of non-zero coefficients surrounding a spatial location corresponding to a transformed data value.

13. A method as claimed in claim 12, wherein the surrounding values utilized are previously entropy decoded values adjacent to a current spatial location corresponding to a transformed data value.

14. A method as claimed in claim 11, wherein said transforming step further includes inverse quantizing transformed portions of said data. 20 15. A method as claimed in 11, wherein said inverse transform includes inverse S- wavelet transforming the data.

16. A method as claimed in claim 15, wherein said data include each of the sub- band components of the wavelet transform which are separately entropy encoded.

18. A method as claimed in claims 16, wherein the lowest frequency sub-band component is separately decoded. i *I 19. A method as claimed in claim 11, wherein said data includes image data describing an image.

20. A method as claimed in claim 11, wherein said magnitude encoding includes encoding the number of leading zeros in transformed data values. 1,

21. An apparatus for compressing data, the apparatus including: transform means for applying a transform to the data to produce transformed data having a series of parts; I *Elm -f -22- entropy encoding means for entropy encoding the magnitude of the transformed data of at least one of said parts based on the magnitude of the surrounding transformed data; and encoding means for separately encoding the value of said transformed data.

22. An apparatus as claimed in claim 21, wherein said entropy encoding means utilizes the number of non-zero coefficients surrounding a spatial location of a corresponding transformed data value.

23. An apparatus as claimed in claim 22, wherein the surrounding values utilized are previously entropy encoded values adjacent to a current spatial location of the corresponding transformed data value.

24. An apparatus as claimed in claim 21, wherein said transform means further includes quantizing means for quantizing transformed portions of said data to integer values. An apparatus as claimed in claim 24, wherein said integer values include a sign bit and a predetermined number of coefficient bits. S*e.

26. An apparatus as claimed in 21, wherein said transform means includes wavelet transform means for wavelet transforming the data.

27. An apparatus as claimed in claim 26, wherein said parts include each of the 25 sub-band components of the wavelet transform which are separately entropy encoded.

28. An apparatus as claimed in claims 26 or 27, wherein the lowest frequency sub- S. band component is separately encoded.

29. An apparatus as claimed in claim 21, wherein said data includes image data describing an image. An apparatus as claimed in claim 21, wherein said magnitude encoding includes encoding the number of leading zeros in transformed data values.

31. An apparatus for decompressing data, the apparatus including; i~WP~ PDD~- BI ~-~arp~-Farsanr: -23- entropy decoder means for entropy decoding a portion of the data to generate the magnitudes of transform data based on the magnitudes of surrounding, previously entropy decoded, transformed data; and decoder means for separately decoding another portion of the data to generate the values of said transformed data; and inverse transform means for applying an inverse transform to the transform data to generate decompressed data.

32. An apparatus as claimed in claim 31, wherein said entropy decoder utilizes the to number of non-zero coefficients surrounding a spatial location corresponding to a transformed data value.

33. An apparatus as claimed in claim 32, wherein the surrounding values utilized are previously entropy decoded values adjacent to a current spatial location corresponding to a transformed data value.

34. An apparatus as claimed in claim 31, wherein said inverse transform means further includes inverse quantization means for inverse quantizing transformed portions S of said data- 2 0

35. An apparatus as claimed in 31, wherein said inverse transform means includes inverse wavelet transform means for inverse wavelet transforming the data.

36. An apparatus as claimed in claim 35, wherein said data include each of the S. 25 sub-band components of the wavelet transform which are separately entropy encoded.

38. An apparatus as claimed in claim 36, wherein the lowest frequency sub-band component is separately decoded. 3 0

39. An apparatus as claimed in claim 31, wherein said data includes image data describing an image. An apparatus as claimed in claim 31, wherein said magnitude encoding includes encoding the number of leading zeros in transformed data values.

41. A computer program product including a computer readable medium having recorded thereon a computer program for compressing data, the computer program product including: B~ 'ti~n -24- transform means for applying a transform to the data to produce transformed data having a series of parts; entropy encoding means for entropy encoding the magnitude of the transformed data of at least one of said parts based on the magnitude of the surrounding transformed data; and encoding means for separately encoding the value of said transformed data.

42. A computer program product including a computer readable medium having recorded thereon a computer program for decompressing data, the computer program product including: entropy decoder means for entropy decoding a portion of the data to generate the magnitudes of transform data based on the magnitudes of surrounding, previously entropy decoded, transformed data; and decoder means for separately decoding another portion of the data to generate is the values of said transformed data; and inverse transform means for applying an inverse transform to the transform data to generate decompressed data. S 43. A method of compressing data, the method including the steps of: 20 applying a transform to the data to produce a plurality of transform coefficients, wherein each transform coefficient is expressible by a code representation including a plurality of symbols; S entropy encoding one of said symbols, not previously entropy coded, of a current transform coefficient based on a context of surrounding symbols; 25 repeating step a predetermined number of times for the current S. transform coefficient; and processing another transform coefficient in accordance with steps and o3 44. A method as claimed in claim 43 wherein said context of surrounding symbols is determined from previously encoded coefficients. A method as claimed in claim 43, wherein the method includes a further step of quantising said transform coefficients. S46. A method as claimed in claim 43, wherein said predetermined number of times is consistent with an encoding of substantially all of the symbols of the current transform coefficient. I

47. A method as claimed in claim 43, wherein said context is determined from an arrangement of surrounding symbols.

48. A method as claimed in claim 47, wherein said surrounding symbols are previously encoded symbols.

49. A method as claimed in claim 48, wherein said context includes a first flag which indicates whether or not a most significant symbol of the current transform 1o coefficient has been encoded. A method as claimed in claim 49, wherein said context includes a second flag which indicates whether or not a most significant symbol, of at least one transform coefficient spatially adjacent to the current transform coefficient, has been encoded.

51. A method of compressing data including the steps of: a) applying a transform to the data to produce a plurality of transform coefficients, wherein each transform coefficient is expressible by a binary code representation having a plurality of bits; 20 b) entropy encoding one of said bits, not previously entropy coded, of a current transform coefficient based on a context of surrounding bits; c) repeating step b) a predetermined number of times the current transform coefficient; and d) processing another transform coefficient in accordance with steps b) i 25 and c).

52. A method as claimed in claim 51,- wherein said context of surrounding bits is determined from previously encoded coefficients.

53. A method as claimed in claim 51, wherein the method includes a further step of quantising said transform coefficients. Si 54. A method as claimed in claim 51, wherein said context of surrounding bits includes information as. to whether or not a most significant bit of the current transform coefficient has been encoded. A method as claimed in claim 51 or 54, wherein said context of surrounding K.o j bits includes information as to whether or not a most significant bit of at least one s 'g~~~nrs~aa *BR~a~l I 26 transform coefficient spatially adjacent, to the current transform coefficient, has been encoded.

56. A method as claimed in claim 51, wherein said transform coefficients are s represented in a bit-plane representation and said surrounding bits are bits in a current bit-plane.

57. A method as claimed in claim 43 or 51, wherein said entropy encoding is performed by an arithmetic coder.

58. A method as claimed 43 or 51, wherein said transform is a Discrete Wavelet Transform.

59. A method of decompressing data, the method including the steps of: entropying decoding said data to generate a symbol of a current transform coefficient based on a context of surrounding symbols; repeating step a predetermined number of times for the current transform; generating another transform coefficient in accordance with steps and 20 and applying an inverse transform to the transform coefficients to produce data. A method as claimed in claim 59, wherein said context of surrounding symbols is determined from previously encoded coefficients.

61. A method as claimed in claim 59, wherein the method includes a further step -j of inverse quantising said transform coefficients.

62. A method as claimed in claim 59, wherein said predetermined number of times is consistent with an decoding of substantially all of the symbols of the current transform coefficient.

63. A method as claimed in claim 59, wherein said context is determined from an arrangement of surrounding symbols.

64. A method as claimed in claim 63, wherein said surrounding symbols are previously decoded symbols. .,1 ~w~sr~BB~g~PBs~aqh8~ -4 -27- A method as claimed in claim 64, wherein said context includes a first flag which indicates whether or not a most significant symbol of the current transform coefficient is encoded or decoded.

66. A method as claimed in claim 66, wherein said context includes a second flag which indicates whether or not a most significant symbol, of at least one transform coefficient spatially adjacent to the current transform coefficient, is encoded or decoded.

67. A method of decompressing data, the method including the steps of: entropying decoding said data to generate a bit of a current transform coefficient based on a context of surrounding bits; repeating step a predetermined number of times for the current transform; generating another transform coefficient in accordance with steps and (b) applying an inverse transform to the transform coefficients to produce data. S68. A method as claimed in claim 67, wherein said context of surrounding bits is determined from previously encoded coefficients. "69. A method as claimed in claim 67, wherein the method includes a further step •of inverse quantising said transform coefficients. S 25 70. A method as claimed in claim 67, wherein said context of surrounding bits includes information as to whether or not a most significant bit of the current transform coefficient is encoded or decoded.

71. A method as claimed in claim 67 or 70, wherein said context of surrounding bits includes information as to whether or not a most significant bit of at least one ;i transform coefficient spatially adjacent, to the current transform coefficient, is encoded S. or decoded.

72. A method as claimed in claim 67, wherein said transform coefficients are represented in a bit-plane representation and said surrounding bits are bits in a current I bit-plane. i 7? •i lle--Pass IPs iB,.L AI~~k -28-

73. A method as claimed in claim 59 or 67, wherein said entropy decoding is performed by an arithmetic coder.

74. A method as claimed 59 or 67, wherein said inverse transform is a inverse Discrete Wavelet Transform. An apparatus for compressing data, the apparatus including: transform means for applying a transform to the data to produce a plurality of transform coefficients, wherein each transform coefficient is expressible by a code representation including a plurality of symbols; entropy encoder means for entropy encoding one of said symbols, not previously entropy coded, of a current transform coefficient based on a context of surrounding symbols; repetition means for repeating the operations of the entropy encoder means a predetermined number of times for the current transform coefficient; and processor means for processing another transform coefficient in accordance with the operations of the entropy encoder means and repetition means.

76. An apparatus as claimed in claim 75, wherein said context of surrounding 20 symbols is determined from previously encoded coefficients.

77. An apparatus as claimed in claim 75, wherein the apparatus further includes S..quantizing means for quantizing said transform coefficients. J 25 78. An apparatus as claimed in claim 75, wherein said predetermined number of times is consistent with an encoding of substantially all of the symbols of the current 11; transform coefficient.

79. An apparatus as claimed in claim 75, wherein said context is determined from an arrangement of surrounding symbols. t 80. An apparatus as claimed in claim 75, wherein said surrounding symbols are previously encoded symbols.

91. An apparatus as claimed in claim 80, wherein said context includes a first flag which indicates whether or not a most significant symbol of the current transform coefficient has been encoded. __'IiE 29

92. An apparatus as claimed in claim 92, wherein said context includes a second j flag which indicates whether or not a most significant symbol, of at least one transform Scoefficient spatially adjacent to the current transform coefficient, has been encoded. 5 93. An apparatus for compressing data including: transform means for applying a transform to the data to produce a plurality of transform coefficients, wherein each transform coefficient is expressible by a binary' code representation having a plurality of bits; entropy encoder means for entropy encoding one of said bits, not previously io entropy coded, of a current transform coefficient based on a context of surrounding bits; repetition means for repeating the operation of the entropy encoder a predetermined number of times for the current transform coefficient; and processor means for processing another transform coefficient in accordance with the operations of the entropy encoder means and repetition means.

94. An apparatus as claimed in claim 93, wherein said context of surrounding bits is determined from previously encoded coefficients.

95. An apparatus as claimed in claim 93, wherein the apparatus includes quantising means for quantising said transform coefficients.

96. An apparatus as claimed in claim 93, wherein said context of surrounding bits includes information as to whether or not a most significant bit of the current transform 25 coefficient has been encoded.

97. An apparatus as claimed in claim 93 or 96, wherein said context of surrounding bits includes information as to whether or not a most significant bit of at least one transform coefficient spatially adjacent, to the current transform coefficient, has been encoded. S_ a98. An apparatus as claimed in claim 93, wherein said transform coefficients are represented in a bit-plane representation and said surrounding bits are bits in a current bit-plane.

99. An apparatus as claimed in claim 75 or 93, wherein said entropy encoding is 1|; performed by an arithmetic coder. i-

100. An apparatus as claimed 75 or 93, wherein said transform is a Discrete Wavelet Transform.

101. An apparatus for decompressing data, the apparatus including: entropy decoder means for entropying decoding said data to generate a symbol of a current transform coefficient based on a context of surrounding symbols; repetition means for repeating the operation of the entropy decoder means a predetermined number of times for the current transform coefficient; generation means for generating another transform coefficient in accordance io with the operations of the entropy decoder means and the repetition means; and inverse transform means for applying an inverse transform to the transform coefficients to produce data.

102. An apparatus as claimed in claim 101 wherein said context of surrounding symbols is determined from previously decoded coefficients.

103. An apparatus as claimed in claim 101, wherein the method includes inverse quantisation means for inverse quantising said transform coefficients. 20 104. An apparatus as claimed in claim 101, wherein said predetermined number of times is consistent with a decoding of substantially all of the symbols of the current transform coefficient.

105. An apparatus as claimed in claim 101, wherein said context is determined from 25 an arrangement of surrounding symbols.

106. An apparatus as claimed in claim 105, wherein said surrounding symbols are previously decoded symbols.

107. An apparatus as claimed in claim 106, wherein said context includes a first flag which indicates whether or not a most significant symbol of the current transform coefficient is encoded or decoded.

108. An apparatus as claimed in claim 107, wherein said context includes a second 35 flag which indicates whether or not a most significant symbol, of at least one transform Scoefficient spatially adjacent to the current transform coefficient, is encoded or encoded. o- -31

109. An apparatus for decompressing data, the apparatus including: entropy decoder means for entropying decoding said data to generate a bit of a current transform coefficient based on a context of surrounding bits; repetition means for repeating the operation of the entropy decoder means a predetermined number of times for the current transform coefficient; generation means for generating another transform coefficient in accordance a with the operations of the entropy decoder means and the repetition means; and inverse transform means for applying an inverse transform to the transform coefficients to produce data.

110. An apparatus as claimed in claim 109, wherein said context of surrounding bits is determined from previously decoded coefficients.

111. An apparatus as claimed in claim 109, wherein the apparatus includes inverse quantisation means for inverse quantising said transform coefficients.

112. An apparatus as claimed in claim 109, wherein said context of surrounding bits includes information as to whether or not a most significant bit of the current transform coefficient is encoded or decoded.

113. An apparatus as claimed in claim 109 or 112, wherein said context of surrounding bits includes information as to whether or not a most significant bit of at .least one transform coefficient spatially adjacent, to the current transform coefficient, is encoded or decoded.

114. An apparatus as claimed in claim 109, wherein said transform coefficients are i represented in a bit-plane representation and said surrounding bits are bits in a current bit-plane. 1 30 115. An apparatus as claimed in claim 101 or 109, wherein said entropy decoding is performed by an arithmetic coder. i 116. An apparatus as claimed 101 or 109, wherein said inverse transform is a inverse Discrete Wavelet Transform.

117. A computer program product including a computer readable medium having y. recorded thereon a computer program for compressing data, the computer program il Iproduct including: B F i- -32- transform means for applying a transform to the data to produce a plurality of transform coefficients, wherein each transform coefficient is expressible by a code Srepresentation including a plurality of symbols; entropy encoder means for entropy encoding one of said symbols, not previously entropy coded, of a current transform coefficient based on a context of surrounding symbols; and repetition means for repeating the operations of the entropy encoder means a predetermined number of times for the current transform coefficient; and processor means for processing another transform coefficient in accordance 1 0 with the operations of the entropy encoder means and repetition means.

118. A computer program product including a computer readable medium having recorded thereon a computer program for compressing data, the computer program product including: transform means for applying a transform to the data to produce a plurality of transform coefficients, wherein each transform coefficient is expressible by a binary code representation having a plurality of bits; entropy encoder means for entropy encoding one of said bits, not previously entropy coded, of a current transform coefficient based on a context of surrounding S. 20 bits; S repetition means for repeating the operation of the entropy encoder a predetermined number of times the current transform coefficient; and m processor means for processing another transform coefficient in accordance with the operations of the entropy encoder means and repetition means. 25

119. A computer program product including a computer readable medium having recorded thereon a computer program for decompressing data, the computer program product including: entropy decoder means for entropying decoding said data to generate a symbol of a current transform coefficient based on a context of surrounding symbols; -f repetition means for repeating the operation of the entropy decoder means a predetermined number of times for the current transform coefficient; r generation means for generating another transform coefficient in accordance with the operations of the entropy decoder means and the repetition means; and S3s inverse transform means for applying an inverse transform to the transform coefficients to produce data. 77- -1 -33

120. A computer program product including a computer readable medium having recorded thereon a computer program for decompressing data, the computer program product including: entropy decoder means for entropying decoding said data to generate a bit of a current transform coefficient based on a context of surrounding bits; repetition means for repeating the operation of the entropy decoder means a predetermined number of times for the current transform coefficient; generation means for generating another transform coefficient in accordance with the operations of the entropy decoder means and the repetition means; and o1 inverse transform means for applying an inverse transform to the transform coefficients to produce data. DATED this TWENTY FOURTH day of SEPTEMBER 1998 Canon Kabushiki Kaisha Canon Information Systems Research Australia Pty Ltd Patent Attorneys for the Applicants SPRUSON FERGUSON I 3