AU613650B2

AU613650B2 - Data encoding

Info

Publication number: AU613650B2
Application number: AU19655/88A
Authority: AU
Inventors: Andrew Peter Heron; David Geoffrey Morrison
Original assignee: British Telecommunications PLC
Current assignee: British Telecommunications PLC
Priority date: 1987-07-09
Filing date: 1988-07-04
Publication date: 1991-08-08
Anticipated expiration: 2008-07-04
Also published as: NO890991L; NO175450B; AU1965588A; NO175450C; NO890991D0

Description

PC u CJ"

V

INT! RNATIONAI, APPI, K A1 ION Kt It hII D (NI~ M~ TiAT! NT 001P1 R A I W I I AT Y (T) (51) internptionpaltetl (71a~siricIali~ 4 (ii0 international Publication N umber: WO 89/ 00796 1104N 7/133 (21) International Application Numnber: P( (22) IMornAwional Filing Dote: 4 Jtuik (31) PrIorlt) Application Number: A I InonioalPbialo ae .inuar% 1984)(26,01 189) 1I d8NtJI19 10 91 87 (74) Agent: LLOYD). BarrN. (,eorge. William: Btii Tec(.om, Intellectualil1iopert) U~nit. M~ (lower Street.

.o n do n W(C I 6 11A (G [3) Des~ignated States: AL M~l, f 1. NO, U'S.

(32) Prloriti Dale: 9 Jul.\ (33) Priority Country., (il, (71) Aglicant for all deognaiud S1tw e( eli IS' BR IT.

LLOS\1 N\I( A IlONS~ Pt l IIII I iIi 11 I)D C OMPANN JOBl 611]. M1 Newgatc Street. Lon.

doni IA 7A (72) Insonlorst and Iniontors/Applicank fha I'S em/h MW(YN DI{d\ iN tj id. C(woffre\ 1613 0(1 If'T~lui (tvo r.'rlvl I11O).i S%1IIh. Sufh'Ik IT'll' 1614# Ill Ile (or K(03,]3; 44 hdihouwe Road. Nvokdlwn Mijrkc,.

lpS%Lh, Suffolk Il6 st It (ouB) Wl'ith intecrnaiall(~a .tarch repo'rt X. a.J. p.

6 APR 1989

AUSTRALIAN

1 3FL-B 989 PATENT OFFICE (54)Tle: DATA ENCODING 0

ADDRESS

COEFFICIENTS

(57) Abstract Values, such as a set or coefficients in a transform-based video coder are to be transmitted sequentially in any one of a number of possible preset orders, The order to be selec'ted is the one that results in the largest number of consecutive zeros (or low values) at the end of the sequence, Values are supplied (in an arbitrary sequenceN with addresses, in parallel to a bank of assessment sections one for each preset order each of which translates the addresses into addresses representing a position in the respective order and records the highest for which the associated value is non-zero, The outputs of the sections are then compared to identify the lowest, and hence the optimum order of transmission,

A

i -4 -1- The present invention relates to an apparatus and method for encoding a set of values eg of transform coefficients in a video coder.

According to one aspect of the invention there is provided an apparatus for encoding a set of values, comprising: assessment means for receiving the values in a first order and having a plurality of sections each arranged in accordance with a respective different predetermined order of the values to examine the values in the said first order, to identify, of those values which meet a predetermined criterion, which value occupies the highest position in the said respective predetermined order, and to produce an output indicating that position; comparator means responsive to the output of the assessment means to identify that producing the lowest indicated position; and means for output of the values in the order corresponding to the identified assessment means.

In another aspect, the invention provides a method for encoding a set of values using an encoding apparatus, comprising: receiving the values in a first order in said apparatus; examining the values in the said first order, to identify, of those values which meet a predetermined criterion, which value occupies the highest position in 4 each of a plurality of predetermined orders of the values and to produce a signal indi.cating that position; comparing the said signals to identify that one of the predetermined orders having the lowest indicated position; and producing, as output from said apparatus, the values in the order thus identified.

V 0 f quot, NN O 89/00796 PCT/GB88/00519 "2- Other optional features of the invention are defined in the sub-claims.

One embodiment of the invention will now be described, by way of example, with reference to the accompanying drawings, in which: Figure 1 is a block diagram of a coder; Figure 2 illustrates a block of transform coefficients; Figure 3 is a block diagram of the sequence selector 4 of figure 1; Figure 4 is a block diagram of the assessment means 12 of figure 3; Figure 5 is a block diagram of the resequencer 5 of figure 1; and Figure 6 is a block diagram of an alternative form of the assessment means 12 of figure 3.

The video coder shown in Figure 1 has a video input 1, and an analogue to digital converter 2. This is followed by a transform coder 3. Each frame of the picture is notionally divided into blocks of picture elements (pixels) and each block subjected to a two-dimensional transform such as the discrete cosine transform (DCT) to produce a block of coefficients. One object of conversion into the transform domain is to effect a reduction in the quantity of data which needs to be encoded for transmission, since the number of bits used for encoding each coefficient can be tailored to the relative contribution of that coefficient to the picture quality.

In particular, coefficients which are insignificantly small or zero need not be transmitted.

Other redundancy reduction techniques such as the use of inter-frame comparison (before or after the transform coder) so that data needs to be transmitted only in respect of blocks which have changed between frames, differential coding, and the use of motion compensation,

I

-3may also be employed if desired. However these, like transform coding itself, are well known in the picture coding field and will not be described further.

The transform coder output is supplied to a sequence selection unit 4, and a resequencer 5 which outputs the coefficients of a block in an order determined by the unit 4 different from that in which they were received.

As previously mentioned, each frame (or field) of the picture is divided into blocks; for example 8 pixel x 8 line blocks may be employed. After transformation, each block gives rise to an 8x8 block of coefficients, as illustrated schematically in Figure 2 with the coefficients numbered (arbitrarily) 0 to 63.

Conventionally (though not necessarily) these are represented as a matrix ordered as to sequency, the upper left coefficient representing the mean level of brightness of the block (the "dc" coefficient) and coefficients increasing in horizontal and vertical sequency as one moves to the right or downwards.

"Sequency" is the equivalent in the case of a discrete transform of frequency in the case of a continuous transform; the higher sequency coefficients carry information about the higher spatial frequency components of the picture. Generally these are smaller than the dc or lower sequency coefficients and commonly are more coarsely quantised before transmission. Some of these indeed may be zero and others may be so small as to be set to zero by a thresholding process.

In order to reduce the number of coefficients that require to be transmitted, it is proposed to vary the sequence in which the coefficients are transmitted, by defining a number perhaps eight of different orders of transmission. Two possible sequences are illustrated by arrowed lines A,B in Figure 2. Once the last non-zero

-J

iWO 89/00796 PC/GB88/00519 K 4coefficient has been sent, transmission (for that block) can be terminated. The purpose of the selector 4 is to determine which order of transmission results in the maximum number of zeros at the end of the sequence, thereby minimising the number of coefficients that have to be sent.

The selector 4 is shown in more detail in Figure 3 and has an input to which a set of 64 coefficients for each block in turn are supplied by the transform coder 3. The processing of a single block will be described, it being understood that processing of subsequent blocks occurs in like manner.

An address generator 11 produces a sequence of 64 unique addresses synchronously with the appearance of the coefficients at the input. The sequence in which they Sappear is immaterial: for the purposes of further description it will be assumed that the generator is a 6-bit binary counter and addresses are the numbers shown in Figure 2. Assessment means 12 serves to ascertain which of the eight possible orders is best suited to the transmission of the particular block. Each address in turn is passed to it, along with the corresponding coefficient, the latter via a circuit 10 which produces a output if the coefficient is non-zero. The assessment means 12 has eight sections 12(1)-12(8), one of which is shown in Figure 4.

The addresses, applied at input 121, pass to a look up table stored in a read-only memory 122. This stores the position, in the relevant one of the eight orders, of each coefficient ie the output of the memory 122 is a translated address. For example, assuming order B of Figure 2, address 10 represents the eighth coefficient and that order and therefore address 10 translates to 7.

L I WO 89/00796 PC/GB88/00519 As each translated address appears at the output of the memory 122 it is loaded into a latch 123 if it is greater than the previous address stored in the latch

AND

the coefficient is non-zero.

For this purpose there is provided a comparator 124 which receives the outputs of the memory 122 and the latch 123 and produces a output if the former is greater than io the latter. The output of the comparator and the output of the circuit 10 of figure 3 drive a load enable input of the latch 123 via an AND gate 127. After the address generator 11 has cycled through its 64 addresses, the content of the latch 123 (which is the output of the section) will be the largest translated address that is associated with a non-zero coefficient.

Each of the sections 12(1)-12(8) is identical except for the content of the look up table 122 which corresponds in each case to a respective one of the eight orders although if the order in which the coefficients were originally supplied is one of the eight, then one section could omit the table 122. When all 64 addresses have been scanned by the address generator, the outputs of the eight sections are compared to determine which is the lowest.

This function is performed by a comparator tree (Figure 3) consisting of six read-only memories 13-19 each of which receives two addresses from assessment section outputs or an earlier such memory and produces as output the lower address. The outputs minl min6 access a further read-only memory 20 to produce a code indicating which of the eight sections provided the lowest address, the address itself appearing at the output of the last stage 19 of the tree.

WO 89/00796 PT/GB188/0519 6 The resequencer 5 is shown in figure 5. (The address generator 11 is shown again for clarity). The coefficient selection operates in two phases, the first of which has been described above. During this first phase, the coefficients are stored in a store 21 with the aid of the address generator 11. In the second phase, the coefficients are read out from the store; for this purpose the address generator 11 again cycles through its sequence, as before, but the addresses generated pass to 0o the store 21 via a read-only memory 22 containing eight translation tables each of which is the reverse of the contents of all eight stores 12(l)-12(8) ie the generated address is interpreted as the position in the required order and the table provides the corresponding store address. Which of the eight tables is used is determined by the output of the read-only memory 20 which is applied to the higher order address lines of the memory 22. In this way the coefficients are read out in the selected order. The output of the comparator 19 is fed to a comparator 23 which produces a "last coefficient" pulse when the address generator reaches the address indicated.

This pulse can be used to prevent the zero coefficients being entered into the output buffer 6 (Figure 1) and hence to the output 7 of the coder.

In practice, for speed of operation, the two phases can be carried out simultaneously, the store 21 actually consisting of two stores one of which is being written to whilst the other is being read out.

0 Figure 6 shows a modified version of the section shown in Figure 4, in which the function of the lookup table 122 and comparator 124 are performed by a read-only memory 128. This may be regarded as having a row address R provided by the address generator 11 and column address C provided by the output of the latch 123. All locations

J/

WO 89/00796 PCT/GB88/00519 -7having a translated row address greater than the column address contain the translated row address; other locations contain the column address.

The buffer 6 may be preceded by other coding arrangements such as a variable length coder.

In order to decode the transmitted sequence, it is necessary for a decoder to be informed of the transmission sequence used, and hence figure 1 shows the output of the sequence selector 4 connected to the output buffer 6, so that the sequence code output by the lookup table 20 of figure 3 is included in the transmitted data.

The number of coefficients sent may be communicated explicitly (by transmitting also the output of the comparator 19) or implicitly as is assumed below by transmitting an end of run code following the last coefficient.

The circuit shown in figure 5 may also be used in a decoder. The received signal would first be preprocessed (not shown) by a variable length decoder or other means appropriate to the method of transmission used. Then the sequence code is stripped by an additional unit 24 (shown in broken lines) and supplied to the lookup table 22. The coefficients are entered into the store 21 by means of the address generator 11 (driven by suitable clock recovery means, not shown) and read out using the part of the lookup table 22 selected by the received sequence code.

They may then be subjected to an inverse transform operation.

It will be appreciated that some or all of the functions of the coder sequence selector 4 and/or the resequencer 5 could if desired be performed by suitably programmed digital processing means.

A typical program listing in pseudo-language is set out below for the sequence selection. The first part of

,J

T~i i L WO $89/00796 PCT/GB88/00519 -8the 'ACTION' segment performs the same function as the assessment means in figure 3 and finds the position of the last non-zero coefficient in each scanning order. The second part examines these positions in turn to find the minimum and performs the same function as the tree structured comparators 13-20 in figure 3. The program assumes 64-element blocks, but any desired block size may of course be used.

SEQUENCE SELECTION PROGRAM

DECLARE:

VALUE (64) TABLE 64) HIGHEST (n) Transform coefficient values Array of order tables for n sequences

SEQ

COEFF

TEMP

BEST

Array to store highest each sequence Sequence counter Coefficient counter Workspace register Workspace register position so far for END DECLARE _1 1 i WO 89/00796 PCT/GB88/00519 -9-

ACTION:

FOR COEFF 1 TO 64 IF VALUE (COEFF) 0 NO ACTION

ELSE

FOR SEQ 1 TO n TEMP TABLE (SEQ, COEFF) IF TEMP> HIGHEST (SEQ) HIGHEST (SEQ) TEMP

ELSE

NO ACTION END IF NEXT SEQ END IF NEXT COEFF BEST 1 TEMP HIGHEST (1) FOR SEQ 2 TO n IF HIGHEST (SEQ) TEMP BEST SEQ TEMP HIGHEST (SEQ)

ELSE

NO ACTION END IF NEXT SEQ END ACTION (Selected Sequence now indicated by BEST and number of coefficients to be sent is in TEMP).

J

10 f

S

S S *5

S*

S

THE CLAIMS DEFINING THE INVENTION ARE AS FOLLOWS:- 1. An apparatus for encoding a set of values, comprising: assessment means for receiving the values in a first order and having a plurality of sections each arranged in accordance with a respective different predetermined order of the values to examine the values in the said first order, to identify, of those values which meet a predetermined criterion, which value occupies the highest position in the said respective predetermined order, and to produce an output indicating that position; comparator means responsive to the output of the assessment means to identify that section having the lowest indicated position; and means for output of the values in the order corresponding to the identified section of the assessment means.

2. An apparatus according to claim 1 further comprising address generating means arranged to generate a reference sequence of addresses and apply them to the assessment means and to the output means.

3. An apparatus according to clai.. 2 in which each section of the assessment means includes a look-up table responsive to the generated addresses to provide translated addresses representing the corresponding positions in the respective predetermined order and a latch for storing one address, and is arranged to store in the latch any translated address which exceeds the current content of the latch in the event that the associated value meets the criterion.

4. An apparatus according to claim 3 in which each section of the assessment means has a comparator for comparing each translated address with the content of the relevant latch.

5. An apparatus according to claim 3 in which, in each section of the assessment means, the look-up table is responsive both to the generated addresses and the address stored in the latch to produce as output the generated address or the address stored in the latch according to I S

S

S. S 'p *D i -,Not, I 11 4 4 4 4..

4, 4 *r '.4.44

S

4 4 i MLZc whether the generated address is respectively greater than or not greater than the address stored in the latch, and the latch is connected so to be loaded with that output whenever the associated value meets the predetermined criterion.

6. An apparatus according to any one of claims 2 to 5 in which the output means comprises a store operable to store the values in locations determined by the generated reference sequence of addresses and to retrieve the values in a sequence determined by a look-up table controlled by the assessment means.

7. An apparatus according to any one of the preceding claims in which the predetermined criterion is that the value is more than a set threshold.

8. An apparatus according to claim 7 in which the predetermined criterion is that the value is non-zero.

9. An apparatus according to any one of the preceding claims in which the output means is arranged to output values only up to the said lowest indicated position.

10. A video coder comprising transform coding means and an apparatus according to any one of the preceding claims for encoding sets of values representing transform coefficients.

11. A method for encoding a set of values using an encoding apparatus, comprising: receiving the values in a first order in said apparatus; examining the values in the said first order, to identify, of those values which meet a predetermined criterion, which value occupies the highest position in each of a plurality of predetermined orders of the values and to produce a signal indicatina that position; comparing the said signals to identify that one of the predetermined orders having the lowest indicated position; and producing, as output from said apparatus, the values in the order thus identified.

I. I

-J

4 4 12 12. A method according to claim 11, in which the predetermined criterion is that the value is more than a set threshold.

13. A method according to claim 12 in which the predetermined criterion is that the value is non-zero.

14. A method according to any one of claims 11, 12 or 13 in which the values are output only up to the said lowest indicated position.

DATED this 22nd day of May 1991 BRITISH TELECOMMUNICATIONS public limited company Attorney: PETER HEATHCOTE Fellow Institute of Patent Attorneys of Australia of SHELSTON WATERS S S i 4, i ft 0* 0 0 *s o a.

0 go*ftf so ft o:

Claims

1. CLASSIFICATION OF SUBJECT MATTER (1I levefol C140111 101101n symbol# 4001Y indIt All) i According to Inteinatigoial P210111 Classification (IPC; of 1o both National Classification and IPC IPC4 H 04 N 7/133 11, FIELDS SEARCHED Minimum Documientation SearchedI Classification tayetom I Classrflcation Symbols Documentation Searched other than Minimum Docuimentation to the Extent that such Documents or# Included In the Fields Searched Ill, DOCUMENTS CONSIDERED TO sE RELEVANT' Category *I Citation of Document, I' with indication. where Appropriate, Of the relevant pseagac Is Relevant to Claim No," A WO, A, 87/02854 (SOCI:ETE FRANCAISE POUR 1,10 LA GESTION DES BREVETS D'APPLICATION NUCLEAIRE BREVATOME) 7 May 1987 see page 21, line 14 page 26, line 28 X 11 A IEEE international conference on 1,10 Communications, 22-25 June 1986, Toronto, CA, volume 1, IEEE, (New York, US), J. Guichard et al.: "Intra and inter- frame transform coding for moving pictures transmission", pages 381-384 see page 383, left-hand column, line 12 right-hand column, line 3 A IEEE Global Telecommunications conference, 1,10 1-4 December 1986, Houston, Texas, volume 1, IEEE, (New York, US), M. Ohta et al.: "Adaptive VWL coding of transform coefficients for sub- primary -ate video transmission" .1 *Special categoric. of cited documentt as IT* later documernt pubilihed after the intiernational flung date A doumet deinig th geera stee o theartwhic Ianot or priority date end not in conflict with the application but cA omnside n te 1 pat tua eo eevne twihI o cited to understand the principle or theo" underflylng the earfle document but published on or after the Intemratonail docum~ of particular relevance, the claimed Invention filng data canwne be esioderod novel of canntot be cenaideried to document which may throw double on priority etaim(o) or aInve an inentve eso" which Is citeod to establish the publication date of Another documenot of particular relevanec; the claimd invention ctton or other special rieaeon (ae specified) cannot be coniered to Involve an Ineftive step when the document referring to an orul disclosure, ue. eahibltlon or decu mmn Ie combineod with on* or mom other such docu- other means ments. such combination belnf, obviouse to a isn -Willed "13" document Publiehed prior to the international filing date but in th s later than the priority date claimed "e document member of the same patent family IV, CERTIFICATION Date of the Actual Completion of the International Search Data of Mailing of tWte International Search &"_o8n 2nd September 1988 1SP18 International Searching Authority Signat C hrz Ca EUROPEAN PATENT OFFICEPh1E r Form PCTIISA 121 (second sheet) (Januay 19") I -1) i4 1i

-2- Intemational ApPlIcation No. PCT/GB 88/00519 III. DOCUMENTS CONSIDERED TO BE RELEVANT (CONTINUED FROM THE SECOND SHEET) Caltegory Citioatio of Document, witn niarlon wnero oppreorials, of the relevant DOssages Relevant to Claim No A pages 271-275 see page 272, left-hand column, line 1 page 273, right-hand column, line 16 IEEE Transaction on Acoustics, Speech, and Signal processing, volume ASSP-32, no. 1, February 1984, IEEE, (New York, US), K.N. Ngan: "Image display techniques using the cosine transform", pages 173-177 see the whole document 1,10 Form PCT ISA210 (extra sheet) (January 9 -1 ANNEX TO THE INTERNATIONAL SEARCH REPORT ON INTERNATIONAL PATENT APPLICATION NO- GB 8800519 SA 22982 Thils annex lists the patent family members relating to the Patent documents cited in the above-mentioned international search report. The members are as contained in the European Patent Office EDP ile on 15/09/88 The European Patent Ofrnce is in no wiay liable for thcse pa~rticulars i"hich are merely given (or the purpose of information, Patent document Publication Patent family Publication cited in search report date member(s) I date W-A- 8702854 07-05-87 FR-A,B 2589020 24-04-87 EP-A- 0247075 02-12-87 JP-T- 63501257 12-05-88 0 0 w For more details about this annbex See Official Journal of the European Patent Office, No. 12/82