AU754544B2 - Wavelet filtering - Google Patents

Description

S&FRef: 541400

AUSTRALIA

PATENTS ACT 1990 COMPLETE SPECIFICATION FOR A STANDARD PATENT

ORIGINAL

Name and Address of Applicant Actual Inventor(s): Address for Service: Canon Kabushiki Kaisha 30-2, Shimomaruko 3-chome, Ohta-ku Tokyo 146 Japan Dominic Yip Spruson Ferguson St Martins Tower,Level 31 Market Street Sydney NSW 2000 Wavelet Filtering Invention Title: ASSOCIATED PROVISIONAL APPLICATION DETAILS [33] Country [31] Applic. No(s) AU PQ6302 [32] Application Date 16 Mar 2000 The following statement is a full description of this invention, including the best method of performing it known to me/us:- 5815c WAVELET FILTERING Technical Field of the Invention The present invention relates generally to wavelet filters and, in particular, to wavelet filtering of digital images.

Background 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.

The heavy quantisation is an example of a lossy technique of image compression. The ooo JPEG standard also provides for the subsequent lossless compression of the transform S. 15 coefficients.

S o.

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.

Typically, data compression using wavelet techniques is a two step process. It comprises, firstly, a transform phase, during which the wavelet transform of the data set is :calculated, and secondly a subsequent coding stage during which the resultant data set from the transform operation is separated into segments which are then coded using a specific coder. In decompression, the reverse occurs, with coded blocks being first decoded, and subsequently the inverse wavelet transform being applied to generate the final decompressed output.

Typically, the wavelet transform operation utilises a number of wavelet filters.

These filters are typically digital filters, such as finite impulse response (FIR filters) which ideally are linear and require a small number of taps. These filters process the digital images in bands and typically require a large external store for storing overlapping pixels, coefficients, and/or intermediate results for processing the next band. The bandwidth to the external store is typically limited, which restricts the rate of reading and writing of overlapping pixels, coefficients, and/or intermediate results to the store.

541400.doc -2- Disclosure of the Invention It is an object of the present invention to substantially overcome, or at least ameliorate, one or more disadvantages of existing wavelet filters.

According to a first aspect of the invention, there is provided a method of wavelet filtering a digital image, the digital image comprising a plurality of pixels arranged in one or more bands of pixels, the method performing the following steps for each one of said one or more bands of said pixels: retrieving a plurality of previous partial results from a first or second storage; inputting a current group of adjacent said pixels; computing coefficients utilising said current group of adjacent pixels and said plurality of previous partial results; outputting said coefficients; and storing a plurality of current partial results in said first storage if said current group is not the last group in the band, in said second •storage if said group is the last group in the band.

According to a second aspect of the invention, there is provided an apparatus for wavelet filtering a digital image, the digital image comprising a plurality of pixels 15 arranged in one or more bands of pixels, the apparatus comprising: means for retrieving, for each one of said one or more bands of pixels, a plurality of previous partial results from a first or second storage; means for inputting, for each one of said one or more bands of pixels, a current group of adjacent said pixels; means for computing coefficients, for each one of said one or more bands of pixels, utilising said current group of adjacent pixels and said plurality of previous partial results; means for outputting, for each one of said one or more bands of pixels, said coefficients; and means for storing a plurality of current partial results in said first storage if said current group is not the last group in the band, in said second storage if said group is the last group in the band.

According to a third aspect of the invention, there is provided a computer readable medium comprising a computer program for wavelet filtering a digital image, the digital image comprising a plurality of pixels arranged in one or more bands of pixels, the computer program comprising: code for retrieving, for each one of said one or more bands of pixels, a plurality of previous partial results from a first or second storage; code for inputting, for each one of said one or more bands of pixels, a current group of adjacent said pixels; code for computing coefficients, for each one of said one or more bands of pixels, utilising said current group of adjacent pixels and said plurality of previous partial results; code for outputting, for each one of said one or more bands of pixels, said coefficients; and code for storing a plurality of current partial results in said first storage 541400.doc if said current group is not the last group in the band, in said second storage if said group is the last group in the band.

According to a fourth aspect of the invention, there is provided a wavelet filter for wavelet filtering a digital image, the digital image comprising a plurality of pixels arranged in one or more bands of pixels, the filter comprising: a first storage for storing a plurality of previous partial results; a second storage for storing a plurality of previous partial results; a controller for selecting the plurality of previous partial results from said first or second storage; a pixel input mechanism for inputting a current group of adjacent said pixels;- a lifting lattice of multiplier and adder units for computing coefficients utilising said current group of adjacent pixels and said selected plurality of previous partial results; output means for outputting said coefficients; and a controller for storing a plurality of current partial results in said first storage if said current group is not the last group in the band, in said second storage if said group is the last group in the band.

Brief Description of the Drawings S: 15 A number of preferred embodiments of the present invention will now be described with reference to the drawings, in which: S"Fig. 1 is a schematic representation of a K-line 5-3 wavelet filter (with K 8) for filtering a digital image in accordance with a first embodiment; Fig. 2 is a schematic representation of the cascading of a wavelet filter of Fig. 1 in 20 time; and °oo* Fig. 3 is a flow diagram of a method for use in wavelet filtering in accordance with a second embodiment; and Fig. 4 is a schematic representation of a general-purpose computer for use in implementing the method of Fig. 3.

Detailed Description including Best Mode Where reference is made in any one or more of the accompanying drawings to steps and/or features, which have the same reference numerals, those steps and/or features have for the purposes of this description the same function(s) or operation(s), unless the contrary intention appears.

The principles of the preferred embodiments described herein have general applicability to the implementation of K-line wavelet filters. However, for ease of explanation, the embodiments has been described with reference to a K-line 5-3 wavelet filter for use in a forward discrete wavelet transform of digital images. However, it will 541400.doc -4be readily evident that the invention is not limited thereto. For instance, the wavelet filter can also be used in K-line 9-7 wavelet filters. Furthermore, it would be apparent to a person skilled in the art, that the wavelet filter can be modified for use in an inverse discrete wavelet transform. Still further, it would be apparent to a person skilled in the art that two such wavelet filters can be combined in known manner to provide a 2dimensional discrete wavelet transform or modified for use in a 2-dimensional inverse discrete wavelet transform. Moreover, the wavelet filter can be used in many different applications. 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, pages 25 to 26. For a discussion of the different applications of wavelets in computer graphics, reference is made to "Wavelets 'i for Computer Graphics", page 5, I. Stollinitz et. al. published 1996 by Morgan Kaufnann Publishers, Inc.

~As mentioned above, the embodiments of the invention can be used in a discrete 15 wavelet filter transform as well as an inverse discrete wavelet transform. For sake of simplicity, the term wavelet transform used herein or variations thereof, is taken to include both a forward and/or an inverse wavelet transform, unless the contrary intention appears. Similarly, the term pixel as used herein or variations thereof, is taken to taken to refer to the original pixels or wavelet coefficients of a digital image, unless the contrary intention appears. Similarly, the term digital image or variations thereof is taken include an original image or a sub-band of that original image, or one or more associated subbands of wavelet transformed coefficients of that original image.

Turning now to Fig. 1, there is shown a schematic representation of a K-line 5-3 wavelet filter (with K 8) for filtering a digital image in one dimension in accordance with a first embodiment. The digital image comprises a plurality of pixels arranged in a plurality of columns in the vertical direction and a plurality of rows in the horizontal direction.

The filter 100 processes one horizontal band of the digital image after another. The digital image comprises one or more such horizontal bands, with the horizontal bands comprising a plurality of said rows of said pixels. The filter 100 initially commences processing the uppermost band of the image, and then processes the next adjacent band of the image, and so on until the lowermost band of the image is processed. Within each band, the filter 200 processes one column of pixels after another. The filter 100 first 541400.doc processes the left column of the band, and then the next adjacent column of the band, and so on until the right column of the band is reached. For ease of understanding, the pixels in each column of a band from the topmost pixel to the lowermost pixel are represented as Xo. X, X2. XK'--I.

Within each column of a band, the filter 100 processes a first group of adjacent pixels (commencing with the topmost group) and then processes a second group of adjacent pixels adjacent the first group.

In this way, the filter 100 performs a one dimensional discrete wavelet transform in the vertical direction. In another embodiment, the filter 100 can also be adapted to perform a one dimensional discrete wavelet transform in the horizontal direction by processing vertical bands one after another in similar fashion. In still another embodiment, the filter can also be adapted to provide a 2-dimensional wavelet transform.

In the latter case, one filter 100 processes one column at a time first, and a second filter 100 after that collects the output of the first filter 100 and processes one row at a time.

The pixels e.g. xo X2, x3, x 4 .xs X6, X7, X8, 9,g Xg x 1 o, X12, X13. X4 and x/ 1 of a column of a band are fed to the filter 100. These pixels are fed in parallel to the wavelet filter 100 via a number 2K (with K 8) of input lines 101. The 16 input lines 101 are connected to eight 2-input multiplexers 103-1, 103-2, 103-3, 103-4, 103-5, 103-6, 103-7, and 103-8.

The multiplexers can select either the bottom eight lines or top eight lines for input of the 20 groups of pixels (xo. xi. x 2 x 3 x4. x5 x 6 and X7) or (xg, x9, xIO. XII, x 12 X13 X1 4 and xis) respectively. The multiplexers 103-1, 103-2, 103-3, 103-4, 103-5, 103-6, 103-7, and 103- 8 are connected to a lifting lattice of multiplier and adder units 102, 104, 106, 108, 110, 112, 114, and 116, which compute the high pass coefficients and low pass coefficients on the top 8 lines (ie. coefficients (xs. x9, xo 0 xI, X12, x 1 3 x14 and xis)) or the bottom 8 lines ie. coefficients (xo. x 1 x2, X3. x4, x 5 x6, and The lifting lattice of multiplier and adder units 102, 104, 106, 108, 110, 112, 114, and 116, perform the following calculations: d, =x2 a(x 2 2 n -1 (1) and, x 2 d, n 2 (2) where d, are all the high pass coefficients, s, all the low pass coefficients, c and P 0.25.

541400.doc -6- For example, the multiplier and adder units 102, 104 106, 108, 110, 112, 114, and 116 perform the following calculations on the selected input pixels (xo xl, x2, 3, x 4, x5 x6, and x7) to compute the high pass coefficients d, and the low pass coefficients s,: unit 102 d o x o (3) unit 104 d, =x 2 a(x +x 3 (4) unit 106 d 2 4 a(x x 5 unit 108 d 3 x 6 a(x x, (6) unit 110 s_ x_ do) (7) unit 112 so x, P(do (8) 0 unit 114 s,=x f(d, (9) unit 116 s =x 5 f(d 2 d) a.

The high pass and low pass coefficients d, and s, are then passed to other discrete 15 wavelet transform circuits or entropy coder (not shown in the figure).

The multiplier and adder units 110 and 102 also accept inputs from either the partial result registers 109 or external buffer 107, depending on the control on the multiplexer 105. Namely, the multiplier and adder units 110 and 102 accept input from the external buffer 107 when computing the first 8 lines and accepts input from the partial S 20 result registers 109 for the last 8 lines.

SA controller 111 controls the storing of a last input pixel and a last high pass coefficient x7 and d 3 to the partial result register 109 or external buffer 107 depending on which lines are being processed. Namely, the last input pixel and the last high pass coefficient of the first 8 lines are stored in the partial result register 109.

However, when the lifting lattice of multiplier and adder units is computing coefficients for the last 8 lines, the last input pixel and last high pass coefficient are stored in the external buffer 107. The last input pixel and last high pass coefficient are required for computation of the low and high pass coefficients of the next adjacent group of pixels within the column.

For example, in the case where the filter 100 is currently processing the pixels (x 8 x 9 xio. Xl. x 1 2 x 1 2 x 14 and XIs), the multiplier and adder units 110 and 102 accept as input the last pixel and coefficient x 7 and d 3 from the partial result registers 109, which were 541400.doc previously stored in the partial result register 109 during the previous computation of the pixels (xo. x l x2. x. X 4 XS. X6. and X7) of the same column and band. In the case where the filter 100 is currently processing the pixels (xo to x 7 the multiplier and adder units 110 and 102 accept as input the last pixel and coefficient xs15 and d 7 from the external buffer 107. These inputs x 1 5 and d 7 were previously stored in the external buffer 107 during the previous computation of the pixels (x 8 to xS15) of the same column but in the last previous band. In the case where there are no previous pixels topmost band) the boundary is symmetrically reflected.

Figure 2 illustrates the cascading of the wavelet filter of Fig. 1 in time. In cycle 0, for the first set of inputs the wavelet filter 100 either symmetrically reflects the pixels near the boundary or reads the required partial results from external buffer 107. The filter 100 then stores the partial results into the partial results register 109, and outputs the oooo, low and high pass coefficients. In cycle 1, the lifting lattice is 'moved' to the next set of input lines, reads input from partial results register 109, and outputs the next low and high 15 pass coefficients. The partial results are stored in the partial results register 109 and external buffer 107.

The first embodiment has been described with respect to a filter 100 having two groups of eight input lines 101 and eight two-input multiplexers 103-1 to 103-8. It would be apparent to a person skilled in the art, that the number of groups can be increased (e.g.

S 20 N 4 groups) with a corresponding modification of the eight multiplexers, (ie. eight Ninput multiplexers). The external buffer 107 is a slow memory storage and requires at least a memory size equivalent to the size of an input pixel and a high pass coefficient times the number of columns. The filter 100 stores in external memory the last input pixel and the last high pass coefficient xS 15 and d 7 of each column of each band for further processing in the next band. On the other hand, the partial results register is a fast local storage and requires only a memory size of an input pixel and high pass coefficient.

The effective value of K' (the width of the band) can be increased by increasing the number of groups of pixels (eg. two groups, K' 2xK 16) or (four groups K' 4xK In the present embodiments, this has the advantage of reducing the number of accesses to the slow external memory. Namely, where there are multiple groups, the last pixel and last high pass coefficient can be stored in the fast local memory, namely the partial results register 109, rather than the slow external memory 107, thus enabling the wavelet transform to be performed more efficiently. Of course, the last pixel and last 541400.doc high pass coefficient of the last group is required to be stored in external memory.

Furthermore, the amount of hardware required to implement the filter is not significantly increased. Still further, the filter has the capability of filtering with different values of K'.

In an alternate embodiment, the multiplexers 103-1, 103-2, 103-3, 103-4, 103-5, 103-6, 103-7, and 103-8 may be dispensed with. In this case, the multiplexing is done by pixel fetching mechanism implicitly. For instance, the pixels are fed to the lifting lattice from a frame buffer by an address generator.

Turning now to Fig. 3, there is shown a method for use in wavelet filtering in one dimension in accordance with a second embodiment. The method comprises in part a sub-procedure 300 for use in the wavelet filtering of a digital image. As mentioned previously, the digital image comprises a plurality of pixels arranged in a plurality of columns in the vertical direction and a plurality of rows in the horizontal direction. The digital image comprises one or more horizontal bands, each of which comprise a plurality S of said rows of pixels. The method processes one horizontal band of the digital image S- 15 after another. The method initially commences processing the uppermost band of the image, and then processes the next adjacent band of the image, and so on until the lowermost band of the image is processed. Within each band, the method processes one column of pixels after another. The method first processes the left column of the band, and then the next adjacent column of the band, and so on until the right column of the 20 band is reached. For ease of understanding, the pixels in each column of a band from the o topmost pixel to the lowermost pixel are represented as xo, xl, x 2 XK'-l.

The method of wavelet filtering calls the sub-procedure 300 for processing each column of a band. Within each column of a band, the sub-procedure 300 processes in turn a first group of K adjacent pixels, a second group of K adjacent pixels adjacent the first group, a third group of K adjacent pixels adjacent the second group, and so on to the last group of pixels. The top and bottom pixel of the first and last group of pixels respectively being located adjacent the boundary of the band.

In this way, the method of wavelet filtering performs a one dimensional discrete wavelet transform in the vertical direction. In another embodiment, the method can also be adapted to perform a one dimensional discrete wavelet transform in the horizontal direction by processing vertical bands one after another in similar fashion. In still another embodiment, the method can also be adapted to provide a 2-dimensional wavelet 541400.doc transform. In the latter case, the method processes one column at a time first, and the method after that collects the output and processes one row at a time.

The sub-procedure commences at step 301 and proceeds to step 302 where a counter M is set to one. As mentioned previously, the sub-procedure processes the pixels within a column of a band. The sub-procedure continues to a loop 306, 308, 310, 312, 314, and 316, where groups of a number K adjacent pixels are retrieved and processed one after another. A first group of K adjacent pixels (nearest the top boundary) is processed during the first pass of the loop 304 to 313, a next group of K adjacent pixels (adjacent the first group) is processed during the second pass of the loop 304 to 313, and so on. The processing of one group of pixels after another is achieved by incrementing the counter M. The sub-procedure terminates once K' pixels have been processed, when M.K Preferably, K 4 and 4 or 8.

After the counter M has been set to one, the sub-procedure proceeds to step 304. In 0 step 304, the method reads the corresponding previous partial results from an external buffer. These previous partial results having being stored in the external buffer during the g processing of the sub-procedure of the same column in the last previous band of pixels.

In the case where there are no previous pixels topmost band) the boundary is symmetrically reflected.

After step 304, the method proceeds to loop 304 to 316. During step 306, the sub-procedure reads the M'th group of K adjacent pixels and then computes the low and high pass coefficients in step 308 in accordance with the equations and described above. In computing these high and low pass coefficients the method takes as input the M'th group of K adjacent pixels. It also accepts input from either the external buffer or a local register which stores the last pixel and the last high pass coefficient. The step 308 outputs to other discrete wavelet transform or entropy encoding procedures (not shown in the figure).

For example, in the case where the sub-procedure is currently processing the pixels (x 8 Xg9, xIO, X11. X12, X13, X14, and XIS), the sub-procedure step 308 accepts as input the last pixel and high pass coefficient x 7 and d 3 from the partial result registers, which were previously stored in the partial result register during the previous computation of the pixels (xo, X. x2, x3. X 4

X

5 X6. and X 7 of the same column and band. In the case where the R sub-procedure is currently processing the pixels (xo to x7), the sub-procedure step 308 541400.doc accepts as input the last pixel and high pass coefficient from the external buffer. These latter inputs were previously stored in the external buffer during the previous computation of the last group of pixels of the same column but in the last previous band. In the case where there are no previous pixels topmost band) the boundary is symmetrically reflected.

Ater step 308, the method proceeds to a decision block 310, where a check is made whether M.K If the decision block returns false, then the method proceeds to step 312, where the last pixel and last high pass coefficient are stored in the local partial results register for use in the computation 308 of the next group of pixels.

In the next step 314, the counter M is incremented and the last pixel and the last high pass coefficient are read 316. The method then returns to step 306 for the next pass of the loop 306 to 316. In the event, the decision block 310 returns true, then the last ooooo pixel and the last high pass coefficient are stored in the external buffer and the subprocedure finishes for the current column of the band.

S 15 As can be seen, the loop can be repeated as many times as desired to increase the value ofK'.

The aforementioned preferred method comprise a particular control flow. There are many other variants of the preferred method which use different control flows without ••go departing the spirit or scope of the invention. Furthermore one or more of the steps of the preferred method may be performed in parallel rather sequential.

The method of Fig. 3 is preferably practiced in dedicated hardware, such as shown in Fig. 1. However, the method of Fig. 3 can also be practiced using a conventional S:general-purpose computer system 400, such as that shown in Fig. 4 wherein the processes S"of Fig. 3 may be implemented as software code, such as an application program executing within the computer system 400. In particular, the steps of method of Fig. 3 are effected by code instructions in the software that are carried out by the computer. This method has particular application in computer systems restricted to large slow external buffers and small fast local memories.

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 541400.doc -11- The use of the computer program readable medium together with the software in the computer system preferably effects an advantageous apparatus in accordance with the embodiments of the invention.

The computer system 400 comprises a computer module 401, input devices such as a keyboard 402 and mouse 403, output devices including a printer 415 and a display device 414. A Modulator-Demodulator (Modem) transceiver device 416 is used by the computer module 401 for communicating to and from a communications network 420, for example connectable via a telephone line421 or other functional medium. The modem 416 can be used to obtain access to the Internet, and other network systems, such as a Local Area Network (LAN) or a Wide Area Network (WAN).

The computer module 401 typically includes at least one processor unit405, a memory unit 406, for example formed from semiconductor random access memory (RAM) and read only memory (ROM), input/output interfaces including a video .':::.interface407, and an 1/O interface413 for the keyboard402 and mouse403 and S. 15 optionally a joystick (not illustrated), and an interface 408 for the modem 416. A storage device 409 is provided and typically includes a hard disk drive 410 and a floppy disk drive 411. A magnetic tape drive (not illustrated) may also be used. A CD-ROM :oooi drive 412 is typically provided as a non-volatile source of data. The components 405 to 413 of the computer module 401, typically communicate via an interconnected bus 404 S 20 and in a manner which results in a conventional mode of operation of the computer system 400 known to those in the relevant art. Examples of computers on which the embodiments can be practised include IBM-PC's and compatibles, Sun Sparcstations or S•alike computer systems evolved therefrom.

Typically, the application program of the preferred embodiment is resident on the hard disk drive410 and read and controlled in its execution by the processor405.

Intermediate storage of the program and any data fetched from the network 420 may be accomplished using the semiconductor memory 406, possibly in concert with the hard disk drive 410. In some instances, the application program may be supplied to the user encoded on a CD-ROM or floppy disk and read via the corresponding drive 412 or 411, or alternatively may be read by the user from the network 420 via the modem device 416.

Still further, the software can also be loaded into the computer system 400 from other computer readable medium including magnetic tape, a ROM or integrated circuit, a magneto-optical disk, a radio or infra-red transmission channel between the computer 541400.doc 12module 401 and another device, a computer readable card such as a PCMCIA card, and the Internet and Intranets including email transmissions and information recorded on websites and the like. The foregoing is merely exemplary of relevant computer readable mediums. Other computer readable mediums may be practiced without departing from the scope and spirit of the invention.

The aforementioned embodiment shown in Fig. 1 can be modified for use in a 5/3 inverse transform wavelet filter. In this modified embodiment, the 5/3 inverse transform wavelet filter calculates the original pixels from the low pass coefficients and high pass coefficients in accordance with the following formulae: x, n=O,1 ,K-1 (11) and, x, =d -a(x2, +2n+ 2 n (12) where a -0.5 and 3 0.25. The multiplier and adder units of Fig. 1 will be modified accordingly to compute the pixels in accordance with equations (11) and This filter S 15 takes as input the coefficients so to sk-i and do to dk-1 and also requires as input the coefficients d.l and x.2 from the previous band, which are stored in the partial register 109 or external buffer 107.

The aforementioned embodiment of Fig. 1 can also be modified without departing from the spirit of the invention to other wavelet filters such as a 9/7 wavelet filters.

In this modified embodiment, the 9/7 forward transform wavelet filter computes the low pass coefficients and high pass coefficients in accordance with the following formulae: d =x 2 n+1 a(X2n +X2n+ 2 n (14) s, =x2n p (d_ n K -1 d, +y(s (16) s, s, n (17) where d, are all the high pass coefficients, s, all the low pass coefficients, d, and are intermediate values, a -1.5861, P -0.052980, y 0.88291, and 5= 0.44351. The multiplier and adder units of Fig. 1 will be modified accordingly to compute the coefficients in accordance with equations (14) to This filter takes as input pixels x, to Xk and also requires as input the pixel xo, the intermediate values d'-i and and the 541400.doc 13 high pass coefficient d-2 from the previous band, which are stored in the partial register 109 or external buffer 107.

In another modified embodiment, the 9/7 inverse transform wavelet filter computes the original pixels from the low pass coefficients and high pass coefficients in accordance with the following formulae: 6(dn, n (18) d' =d (s 2 (19) x 2 x 2 n+1 -a(x 2 +X2n2) n (21) where d, are all the high pass coefficients, s, all the low pass coefficients, and are intermediate values, a -1.5861, P -0.052980, y 0.88291, and 6= 0.44351. The multiplier and adder units of Fig. 1 will be modified accordingly to compute the pixels in accordance with equations (18) to This filter takes as input coefficients so to Sk-1 and do to dk-1 and also requires as input the pixel x-4, the intermediate values d'.

2 and and 15 the high pass coefficient d.l from the previous band which are stored in the partial register 109 or external buffer 107.

The aforementioned embodiment shown in Fig. 3 can also be modified in similar fashion to other wavelet filters, such as a 5/3 inverse transform wavelet filter or 9/7 wavelet filters.

Industrial Applicability It is apparent from the above that the embodiment(s) of the invention are applicable to the computer and data processing industries. In particular, the embodiments have application in digital data compression of digital images and videos.

The foregoing describes only one embodiment/some embodiments of the present invention, and modifications and/or changes can be made thereto without departing from the scope and spirit of the invention, the embodiment(s) being illustrative and not restrictive.

In the context of this specification, the word "comprising" means "including principally but not necessarily solely" or "having" or "including" and not "consisting only of". Variations of the word comprising, such as "comprise" and "comprises" have corresponding meanings.

541400.doc

Claims (9)

1. A method of wavelet filtering a digital image, the digital image comprising a plurality of pixels arranged in one or more bands of pixels, the method performing the following steps for each one of said one or more bands of said pixels: retrieving a plurality of previous partial results from a first or second storage; inputting a current group of adjacent said pixels; computing coefficients utilising said current group of adjacent pixels and said plurality of previous partial results; outputting said coefficients; and storing a plurality of current partial results in said first storage if said current group is not the last group in the band, in said second storage if said group is the last group in ooooo too the band. t 15

2. A method as claimed in claim 1, wherein said first storage is a local storage. to 0.

3. A method as claimed in claim 1 or 2, wherein said second storage is a remote storage. 20

4. A method as claimed in claim 1, wherein said digital image is an original image and said wavelet filtering performs a forward wavelet transform. t.

A method as claimed in claim 1, wherein said digital image is a sub-band of an original image and said pixels are coefficients and said wavelet filtering performs a forward wavelet transform.

6 A method as claimed in claim 1, wherein said digital image is one or more associated sub-bands of an original image, and said pixels are coefficients and said wavelet filtering performs an inverse wavelet transform.

7. A method as claimed in claim 1, wherein said method further comprises: repeating said retrieving, inputting, computing, outputting, and storing steps in sequence a plurality of times.

541400.doc

8. A method as claimed in claim 4 or 5, wherein said computing step is calculated in accordance with a 5/3 forward wavelet transform.

9. A method as claimed in claim 6, wherein said computing step is calculated in accordance with a 5/3 inverse wavelet transform. A method as claimed in claim 4 or 5, wherein said computing step is calculated in accordance with a 9/7 forward wavelet transform. 11. A method as claimed in claim 6, wherein said computing step is calculated in :accordance with a 9/7 inverse wavelet transform. 12. A method as claimed in claim 2, wherein said plurality of partial results stored in 15 said local storage comprises a pixel and a coefficient. 13. A method as claimed in claim 2, wherein said plurality of partial results stored in said local storage comprises a pixel, intermediate values, and a coefficient. 14. Apparatus for wavelet filtering a digital image, the digital image comprising a plurality of pixels arranged in one or more bands of pixels, the apparatus comprising: means for retrieving, for each one of said one or more bands of pixels, a plurality of previous partial results from a first or second storage; means for inputting, for each one of said one or more bands of pixels, a current group of adjacent said pixels; means for computing coefficients, for each one of said one or more bands of pixels, utilising said current group of adjacent pixels and said plurality of previous partial results; means for outputting, for each one of said one or more bands of pixels, said coefficients; and means for storing a plurality of current partial results in said first storage if said current group is not the last group in the band, in said second storage if said group is the last group in the band. 541400.doc 16- Apparatus as claimed in claim 14, wherein said first storage is a local storage. 16. Apparatus as claimed in claim 14 or 15, wherein said second storage is a remote storage. 17. A computer readable medium comprising a computer program for wavelet filtering a digital image, the digital image comprising a plurality of pixels arranged in one or more bands of pixels, the computer program comprising: code for retrieving, for each one of said one or more bands of pixels, a plurality of previous partial results from a first or second storage; code for inputting, for each one of said one or more bands of pixels, a current group of adjacent said pixels; code for computing coefficients, for each one of said one or more bands of pixels, utilising said current group of adjacent pixels and said plurality of previous partial results; code for outputting, for each one of said one or more bands of pixels, said coefficients; and code for storing a plurality of current partial results in said first storage if said current group is not the last group in the band, in said second storage if said group is the last group in the band. 18. A computer readable medium as claimed in claim 17, wherein said first storage is a local storage. 0.0 000.0: 0 19. A computer readable medium as claimed in claim 17 or 18, wherein said second storage is a remote storage. A wavelet filter for wavelet filtering a digital image, the digital image comprising a plurality of pixels arranged in one or more bands of pixels, the filter comprising: a first storage for storing a plurality of previous partial results; a second storage for storing a plurality of previous partial results; a controller for selecting the plurality of previous partial results from said first or second storage; a pixel input mechanism for inputting a current group of adjacent said pixels; 541400.doc -17- a lifting lattice of multiplier and adder units for computing coefficients utilising said current group of adjacent pixels and said selected plurality of previous partial results; output means for outputting said coefficients; and a controller for storing a plurality of current partial results in said first storage if said current group is not the last group in the band, in said second storage if said group is the last group in the band. 21. A wavelet filter as claimed in claim 20, wherein said first storage is a local storage. 22. A wavelet filter as claimed in claim 20 or 21, wherein said second storage is a remote storage. eeee 23. A wavelet filter as claimed in claim 20, wherein said pixel input means comprises a I plurality of multiplexers for selecting in turn a group of pixels as said current group. 24. A wavelet filter as claimed in claim 20, wherein said digital image is an original image and said wavelet filtering performs a forward wavelet transform. A wavelet filter as claimed in claim 20, wherein said digital image is a sub-band of 20 an original image and said pixels are coefficients and said wavelet filtering performs a forward wavelet transform. 26. A wavelet filter as claimed in claim 20, wherein said digital image is one or more associated sub-bands of an original image, and said pixels are coefficients and said wavelet filtering performs an inverse wavelet transform. 27. A wavelet filter substantially as described herein with reference to the accompanying drawings. 28. A method of wavelet filtering, the method substantially as described herein with reference to the accompanying drawings. 541400.doc Is- 29. A computer readable medium comprising a computer program, the computer program substantially as described herein with reference to the accompanying drawings. Dated 15 March, 2001 CANON KABUSHIKI KAISHA Patent Attorneys for the Applicant SPRUSON FERGUSON 0 0ee 00 .0. 0. a 0 0a00 541400.doc