WO2006032167A1

WO2006032167A1 - Video encoding/decoding method

Info

Publication number: WO2006032167A1
Application number: PCT/CN2004/001094
Authority: WO
Inventors: Lianhuan Xiong
Original assignee: Huawei Technologies Co., Ltd.
Priority date: 2004-09-24
Filing date: 2004-09-24
Publication date: 2006-03-30

Abstract

The present invention discloses a video encoding method. The method comprises the steps of firstly determining the code number (code num) corresponding to the code word to be encoded and transmitted; then calculating M and INFO based on the code number according to the rule that M is smaller than INFO, wherein M is the number of bits of the prefix code before the identifier code and N is the number of bits of the information code after the identifier code; finally determining the prefix code based on the number M and encoding with the prefix code, the identifier code and the information code. The invention further discloses a video decoding method corresponding to the above video coding method. According to the invention, the information bits account for larger proportion in the whole encoding data, thus increases the encoding efficiency and reduces the error rate of the prefix code. Therefore the present inventional entropy encoding method possesses a better anti-error performance compared to the UVLC entropy encoding method in which the number of bits is symmetric.

Description

Video coding and decoding method

The invention relates to an information encoding and decoding method.

Background technique

In the field of computer and communication, information is encoded and transmitted. The anti-interference ability of an encoding method directly affects the quality of information transmission. The existing video image coding compression standards MPEG-1, MPEG-2, MPEG-4, H.261, Η·263, H.264, etc. are all based on discrete digital cosine transform (DCT), in which H.264 and The MPEG-4 AVC standard adopts an integer transform method whose performance is close to DCT. The general processing procedure is as follows: When the image is transmitted, the sender first inputs the video image, and the image is divided into specific lattices for motion estimation and motion compensation. Then performing DCT transform and quantization processing, entropy encoding the quantized image data, forming the encoded data stream to be transmitted to the receiving end, performing entropy decoding, inverse quantization processing, motion compensation of DCT inverse transform, etc. by the receiving end The operation obtains the decoded video image. In the above process, the entropy coding method mainly includes variable length coding and arithmetic coding methods. For example, in the H.264 standard, there are three entropy coding methods: a universal variable length coding method (UVLC), and a context-based relationship (context) - bas ed ) variable length coding method (CAVLC ) and context-based arithmetic coding method (CABAC). Among them, the basic idea of the UVLC entropy coding method adopted by the H.264 standard is: For the codeword to be transmitted, its structural form is:

[M zeros] [1] [INFO] or [M ones] [0] [INFO]; where the middle "1" or "0" is the identifier, and the [M zeros] or [M ones] in front of the identifier is The leading character, [INF0] after the identifier is an information character.

In the above structure, it is assumed that "1" is used as the identifier, the number of "0" in front of "1" is the same as the number of bits in the following information. In the H.264 standard, the codeword and the codeword number (code_n draw ) with - corresponding relationship, refer to the following table 1:

Table 1 Correspondence between code_num and Codeword code-iium Codeword

0 1

1 010 2 Oi l

3 00100

4 00101

5 00110

6 00111

7 0001000

8 0001001

9 0001010

10 0001011

11 0001100

12 0001101

1 3 0001110

14 0001111

15 000010000

16 000010001 On the encoding side, for the given codeword sequence code_num value, the number of "0"s before the identifier "1" and the information INFO value after the identifier "1" are calculated by the following method: M = l og ₂ (code—n leg+1)

INFO = code-num + I - 2

At the decoding end, the codeword sequence code_num value is decoded by the following method:

(1) Read the identifier " and the preceding "0" symbol, and set the number of the read symbols "0" to M.

(2) Read the M-bit information following the identifier " 1 " and set its value to INF0.

(3) code_num = 2 + INFO - 1.

Regardless of whether it is the encoding end or the decoding end, if the code word number code_n is 0, both INFO and M are 0, and vice versa.

The above code words have the characteristics of digit symmetry, that is, the number of "0"s in front of the identifier "1" is as many as the number of bits of information behind, and the information bits occupy less than 50% of the entire codeword. Since for a codeword, if the codeword preceding the identifier "1" is in error, it cannot be decoded correctly. Therefore, the excessive number of "0" before the identifier "1" causes the UVLC entropy coding to resist error performance and Reduced coding efficiency. Summary of the invention

In order to improve the coding efficiency and improve the error resistance performance based on the UVLC entropy coding method in the video coding process, the present invention proposes a video coding method and a corresponding video decoding method thereof, and uses the above method to encode information, in front of the identifier. The leading and subsequent information characters are not the same number of bits, and the information bits occupy more bits.

To this end, the video coding method provided by the present invention includes:

Determining the codeword number corresponding to the information that needs to be encoded and transmitted code_mim;

Calculating, according to the codeword number, a bit number M of the preamble preceding the identifier and an information symbol INFO following the identifier, such that M is smaller than the number of bits of the information symbol INFO;

The preamble is determined based on the number of bits M, and the preamble, the identifier, and the information symbol INFO are combined to form an encoded codeword.

In the specific calculation, the number of bits M of the preamble preceding the identifier and the information symbol INFO after the identifier are calculated according to the codeword number according to the following formula:

M = log ₂ (code_num+ 2'! ) _ n;

INFO = code 1 + 2" one 2 ⁺ ";

Where n is the absolute value of the difference between the number of bits M of the preamble preceding the identifier and the information INFO after the identifier, n > 1.

Based on the above calculation, the method further includes: calculating a codeword length Length according to the following formula to facilitate encoding and decoding:

Length = 2M+n+l.

The video decoding method corresponding to the calculation M and INFO modes adopted by the above video coding method includes:

Read the number of preambles in front of the identifier M;

Read the M + n bit information after the identifier, and make it the value INF0;

The code word number _C is calculated according to the following formula. De_num _:

Code— num = 2^ ⁺ " + INFO one 2";

The information to be encoded and transmitted is determined based on the obtained codeword number _CO d _e _num.

In the specific calculation, the number of bits M of the leading character before the identifier and the information symbol INFO after the identifier may also be calculated according to the code word number according to the following formula:

If code_num is less than 2"' -1, Bay ¹ J, M = log ₂ (code_num+ 1 );

INFO = code—num +1— ;

Otherwise:

M = log2(code_num+ 2"' ⁺ "a 2"' + 丄) - ⁿ ;

M+n

INFO = code-num + 2"' ⁺ " - 2'" + 1

Where m > 0, η 1; m is the number of bits in the starting preamble with the number of digits asymmetry in front of the identifier, n is the number of bits M of the leading character before the identifier and the information following the identifier The absolute value of the INFO digit difference.

If code_num is less than 2"' - 1, Bay U,

Length = 2M +1;

Otherwise,

Length = 2M+n+l

(1) reading the number of preambles preceding the identifier M;

(2) If M < m, then

Read the M-bit information after the identifier, and make it the value of INFO;

Otherwise,

(3) If M < m, Bay' J,

Code— 2 ^M + INFO -1;

Otherwise,

Code—num = INFO - 2'" ⁺ " + 2'" a 2^"

The information to be encoded and transmitted is determined according to the obtained code word number code_imm. For the above method, the identifier is binary "1" or "0", when the identifier is ", the preamble is M bit "0", when the identifier is "0" When the leader is M bit "1"

Since the UVLC-based entropy coding described in the present invention has the characteristics of a bit number asymmetry, The number of preceding preamble symbols and the number of information bits after the identifier may be different. The number of symbols of the preamble is less than the number of bits of information after the identifier, so that the information bits occupy more of the entire encoded data. The large specific gravity, which not only makes the invention have higher coding efficiency, but also reduces the probability of leading errors, and therefore, the entropy coding method of the present invention has better performance than the bit-symmetric UVLC entropy coding method. Anti-error performance.

detailed description

The invention proposes a new UVLC-based entropy coding method in the video coding process. If the identifier uses "1", the number of "0" before the identifier "1" and the information bit after "1" The number can be different. For example, the number of "0"s before the identifier "1" 匕1 is less than 1 digit. The following discussion is mainly for this case. The basic idea is: Use "1" as the identifier. In all codewords "codeword", the number of "0" before "1" is less than the number of bits after the "1" is less than n digits, or part of the codeword" In codeword", the number of "0" before "1" is less than the number of bits after "1". Since for a codeword, if the codeword before the identifier "1" is wrong, it cannot be decoded correctly. Therefore, reducing the number of "0" before the identifier "1" can reduce the probability of error, and thus has better error resistance performance. If "0" is used as the identifier, the symbol in front of the identifier is "1". As with other processes, the following analysis is all about using "as an identifier." The present invention can assume an identifier for all codeword cases with a number of bits of asymmetry.

The number of "0" before "1" is less than the number of bits after the "1", and has the following general description:

The structure of the codeword "Codeword" is:

[M zeros] [1] [INFO] ; Based on the above structural form, the first aspect of the present invention is that, at the encoding end, all code words have a bit number asymmetry characteristic. First, determine the codeword number code_num corresponding to the codeword that needs to be encoded and transmitted. For the given codeword sequence code_num value, calculate the number M and the identifier of the "0" before the identifier "1" by the following side. The information INFO value after "1":

M = log2 (code— num+2") - n;

Length = 2M+n+l;

INFO = code one +2" - 2 " ;

Where n is the absolute value of the difference between the number of bits M of the preamble preceding the identifier and the information INFO after the identifier, and η > 1 ;

Finally, the preamble, i.e., M bit "0", is determined according to the number of bits M, and the preamble, the identifier "1", and the information symbol INF0 are encoded.

The length of the code length is to facilitate the encoding and decoding.

The entropy decoding method corresponding to the above entropy coding method includes the following steps: first reading the preamble bit number M in front of the identifier from the received code stream;

Calculate the codeword number code_num according to the following formula:

Code one 诵 = 2 ^M+ " + ITM -2";

Finally, the information to be encoded and transmitted is determined according to the code word number code_mim.

The following describes an embodiment of the above scheme with the code word "Codeword" of the structure [M zeros] [1] [INFO]:

Example 1: n=l. At the encoder, the code word for a given number code- num value _by; the following method to calculate the identifier "in front of" 0 "and the number M of the identifier" 1 "information INF0 following values:

M = log2 (code-num+2) - 1.

Length = 2 (M+l) .

INFO = code-num +2 - 2

At the decoding end, the codeword sequence code-num value is decoded by the following method:

(1) Read the identifier "1" and the preceding "0" symbol, and set the number of the read symbols "0" to (2) Read the M + 1 bit information following the identifier "1" and let it be INF0.

(3) code— num = 2^ + ^INF0 - ²

The correspondence between the code word number code_num and the code word Codeword thus obtained is shown in Table 2.

Table 2 Code word number code_ num and code word Codeword correspondence

Code— num Codeword (H. 264 standard) Codeword (invention)

0 1 10

1 010 11

2 Oi l 01 00

3 00100 0101

4 00101 0110

5 0011 0 0111

6 00111 001000

7 0001000 001001

8 0001001 001010

9 0001010 001011

10 0001011 001100

11 0001100 001101

12 00011 01 001110

1 3 0001110 001111

14 0001111 00010000

15 000010000 00010001

16 000010001 00010010

17 000010010 00010011

18 000010011 000101 00

19 000010100 00010101

20 000010101 00010110

21 000010110 00010111

22 000010111 00011000 23 000011000 00011001

24 000011001 00011010

25 000011010 00011011

26 000011011 00011100

27 000011100 00011101

28 000011101 00011110

29 000011110 00011111

30 000011111 0000100000

31 00000100000 0000100001

32 00000100001 0000100010 Example 2: n=2. At the encoding end, for a given codeword sequence code-num value, the number of "0" preceding the identifier "1" and the information after the identifier "1" are calculated by the following method:

M = log2 (code-iium+4) - 2.

Length = 2 (M+D + l.

INFO = code一诵 +4 - At the decoding end, the codeword sequence code-num value is decoded by the following method:

(1) Read the identifier "1" and the preceding "0" symbol, and set the number of the read symbols "0" to 1.

(2) Read the M + 2 bit information following the identifier "1" and set its value to INF0.

(3) code-num = 2^ + ^{I NF0-4}

The correspondence between the codeword sequence code_num and the codeword Codeword is shown in Table 3.

Table 3 Code word number code-num and code word Codeword correspondence code-num Codeword (H. 264 standard) Codeword (present invention)

0 1 100

1 010 101

2 011 110

3 00100 111 /D/:/ O 60S003⁄4I>d L9s900AV

ΟΟ ΐθθΟΐθΐΟ

s

9

∞ s

ΐΟΠΟ ΟΟΐΙΟΟΟ 6

Ττπο ποιοοο οι

Ππο οοποοο π

000S00 ΐοποοο n

ο οπτοοο n 2

{nosoo πποοο οπο<ΐ§ ΙΟΟΟΐΟΟΟΟ 9ΐ

ΐΞΞΟΟ ΟΐΟΟΟΟΙΟΟ U

Πποΐθο ποοΐοοοο 8ΐ ^ οοοποο §010000

ο ΐοποο 0TIS0000 u

Ϊ́ΐοποο τποοοΐοο ιι

Οοποο οοοποοοο ΐοπΐοο son

Ππποο ποποοοο 9ί

ΟΟΐΟΟΐΟΟΟ ΐΟΟΟΟΐΟΟΟΟΟ Ιί According to a second aspect of the present invention, starting from a certain codeword number, the following codewords have the characteristics of asymmetry of the number of bits, and the preceding codewords have the characteristics of being symmetrical. At the encoding end, first determine the codeword sequence code_num corresponding to the information that needs to be encoded and transmitted; for the given codeword sequence code_num value, calculate the number of "0"s in front of the identifier "1" by the following method M and the identifier "1" after the information INFO value:

If (code-num < ^ ⁿ -l) , Bay 'J,

M = log2 (code_num+l);

Length=2M+l ;

INFO = code - face +1 - 2^;

Otherwise,

M = log2 (code_num+2" ^!+ " - 2'" + 1) - n;

Length = 2M+n+l ;

INFO = code one +2'" ⁺ " - 2"' + 1 - 2^".

Where m > Q , n > l ; m is the number of bits in the starting preamble with the number of digits asymmetry in front of the identifier, n is the number of bits M of the leading character before the identifier and the identifier The absolute value of the difference between the digits of the INF0 symbol; that is, the codeword with the number of "0" in front of the identifier "1" at the beginning of the codeword has a bit-symmetric characteristic, and the identifier "1" is preceded by The number of "0" is less than the number of bits following "1".

Finally, the identifier is determined, and the number of leading symbols "0" is determined based on the number of bits M, and the leading symbols, i.e., M "0", the identifier "1", and the information symbol INF0 are encoded.

The above code length Length is calculated to facilitate the encoding and decoding.

The entropy decoding method corresponding to the entropy coding method described in the second scheme above includes the following steps:

At the decoding end, (1) reading the number of preambles M in front of the identifier;

(2) If M < m, then

Read the M bit information after the identifier, and make it the value INF0;

Otherwise,

Read the M + n bit information after the identifier, and make it the value INF0; (3) If M < m, then

Code—Proposal = 2 ^{M + INF0} - ¹

Otherwise

Code one = INFO - + 2"' - 1 + 2^ ⁺ "

The following is an example of the above-mentioned second scheme (described in conjunction with the PASCAL language) in the code word "Codeword" of the structure [M zeros] [1] [INFO]:

Example 3: n=l, m = l. At the encoding end, for a given codeword sequence code-num value, the number M of "0" before the identifier "1" and the information INFO value after the identifier "1" are calculated by the following method:

If (code-num == 0)

{M = 0, INFO=0, Length=l }

El se

{M = log2 (code_num+3) one 1.

Length = 2 (M+l) .

INFO = code-num +3 - 2

(1) Read the identifier "1" and the preceding "0" symbol, and set the number of the read symbols "0" to ^

(2) if (M≠0)

Read the M + 1 bit information after the identifier "1", and make it the value INF0

(3) if (M==0)

Code_num = 0

El se

Code-num = 2 ⁺ I FO - 3;

The correspondence between the codeword sequence code-num and the codeword Codeword is shown in Table 4.

Table 4 Code word number code_num and code word Codeword correspondence code-num Codeword (H. 264 standard) Codeword (present invention)

0 1 1

1 010 0100 30llllllooool

29 ggllllo〇1§000

00000100000 0000100010

00000100001 0000100011 Example 3: n=l , m = 2. At the encoding end, for a given codeword sequence code_num value, the number M of the "0" preceding the identifier "1" and the information INFO after the identifier "1" are calculated by the following method:

If (code_num <=2)

{M = log2 (code_ num+l)

Length=2M+l

M.

INFO = code—dish m +1 2

El se

{M = log2 (code_num+5) - 1

Length=2 (M+l)

INFO = code— num +5 - 2"

(1) Read the identifier "1" and the preceding "0" symbol, and set the number of the read symbols "0" to M.

(2) if (M<=1)

Read the M-bit information after the identifier "1" and let it be INF0 el se

Read the M + l bit information after the identifier "1" and make the value INF0 ₍

(3) if (M<=1)

Code_num = 2 ^M + INFO - 1

El se

Code-num = 2

The correspondence between the code word number code_num and the code word Codeword is shown in Table 5.

Table 5 code word number code-num and code word codeword correspondence code-num Codeword (H. 264 standard) Codeword (present invention) 29000011110 goolooolo

28s tglolll Issllool

27 gglooll §§010000

30 000011111 0000100011

31 00000100000 0000100100

32 00000100001 0000100101 The above description is only a few specific embodiments of the present invention, and it should be noted that those skilled in the art can make other various kinds without departing from the method of the present invention. An implementation that suits your needs.

Claims

Rights request

A video encoding method, comprising:

Determining the codeword number corresponding to the information that needs to be encoded and transmitted codejrmm;

The video encoding method according to claim 1, wherein the number of bits M of the preamble preceding the identifier and the information symbol INFO following the identifier are calculated according to the codeword number according to the following formula:

M = log ₂ (code_num+ 2" )— n;

INFO = code—mim +2"― 2^ ⁺ ";

Where η is the absolute value of the difference between the number of bits M of the preamble preceding the identifier and the number of bits of the information INFO following the identifier, η > 1.

The video encoding method according to claim 2, wherein the method further comprises: calculating a codeword length Length according to the following formula to facilitate encoding and decoding:

Length = 2M+n+lo

4. A video decoding method according to the video encoding method of claim 2, comprising:

Read the number of preambles in front of the identifier M;

The code word number _C is calculated according to the following formula. De_ _nU m _:

Code side = 2 ^M+ " + INFO ~2";

The information to be encoded and transmitted is determined according to the obtained codeword number code_mim.

If code_num is less than 2'" -1, Bay ¹ J.,

M = log ₂ (code_num+ 1 );

INFO = code— nrnn +1— 2 ; otherwise:

M = log2 (code-num+2'" ⁺ "— 2"' +l) ― n;

INFO = code side +2"' ⁺ "― 2"' + 1 1 2^ ⁺ ";

Where m > 0, η > 1; m is the number of bits in the starting preamble with the number of digits asymmetry before the identifier, n is the number of bits M of the leading character before the identifier and the identifier The absolute value of the information INFO digit difference.

The video encoding method according to claim 5, wherein the method further comprises: calculating a codeword length Length according to the following formula to facilitate encoding and decoding:

If code a num is less than 2'" - 1, Bay ¹ J,

Length = 2M +1;

Otherwise,

Length = 2M+n+l.

7. A video decoding method according to the video encoding method of claim 5, comprising:

(1) reading the number of preambles preceding the identifier M;

(2) If M < m, then

Read the M-bit information after the identifier, and make it the value of INFO;

Otherwise, - read the M + n bit information after the identifier, and make it the value of INFO;

(3) If M < m, then

Code _num = ' f + INFO -1;

Otherwise,

Code one discussion = INFO - 2"' ⁺ " + 2'" - ¹ + 2 ⁺ " °

The information to be encoded and transmitted is determined according to the obtained codeword number code_num.

The video encoding method according to claim 1, 2, 3, 4, 5, 6 or 7, wherein the identifier is binary "Γ or "0" when the identifier is " The preamble is M bit "0", and when the identifier is "0", the preamble is M bit