Detailed description of the H.264 cavlc (encoding and decoding) Process

Http://hi.baidu.com/hero_xu_2006/blog/category/h.264%D1%A7%CF%B0

You can see that the cavlc is coming at 264, and paste the information here:

Encoding Process:
Assume there is a 4*4 data block (change, which is sent to entropy encoding after quantification)
{
0, 3,-1, 0,
0,-1, 1, 0,
1, 0, 0, 0,
0, 0, 0, 0
}
Data re-arrangement:,-1 ......

1) Initial Value Setting:
Number of non-zero coefficients (totalcoeffs) = 5;
Trailingones = 3;
The number of zeros before the last non-zero coefficient (total_zeros) = 3;
Variable NC = 1;
(Note: The NC value is determined: The DC coefficient of the color NC =-1; the NC value of other coefficient types is based on the number of non-zero coefficients (NA) of 4*4 blocks on the left of the current block) the number of non-zero coefficients (NB) of 4*4 blocks in the current block is obtained. For details, see table 6.10 of bihoujieshu p120)
Suffixlength = 0;
I = totalcoeffs = 5; (reverse encoding)

2) coeff_token:
Query criteria (bs iso/IEC 14496-) Table 9-5, available:
If (totalcoeffs = 5 & trailingones = 3 & 0 <= NC <2)
Coeff_token = 0000 100;
Code output = 0000 100; 

3) encode all trailingones characters:
In reverse order, the symbols of the three backend coefficients are + (0),-(1), and-(1 );
That is:
Trailingone sign [I --] = 0;
Trailingone sign [I --] = 1;
Trailingone sign [I --] = 1;
Code output = 0000 100 0 11;
4) non-zero coefficient amplitude levels except tail coefficient: (this example is not very detailed, and there is a small error)
The process is as follows:
(1) convert signed level [I] into a signed levelcode;
If level [I] is positive, levelcode = (level [I] <1)-2;
If level [I] is negative, levelcode =-(level [I] <1)-1;
(2) Calculate level_prefix: level_prefix = levelcode/(1 <suffixlength );
The corresponding bit string is available in the 9-6 table;
(3) Calculate level_suffix: level_suffix = levelcode % (1 <suffixlength );
(4) determine the suffix Length Based on the suffixlength value;
(5) suffixlength updata:
If (suffixlength = 0)
Suffixlength ++;
Else if (levelcode> (3 <suffixLength-1) & suffixlength <6)

Note: If the value is greater than the preset value, suffixlength ++ is used;
Suffixlength ++;

Back in the example, level [I --] = 1; (I = 1)
Levelcode = 0; level_prefix = 0;
Query Table 9-6. The corresponding Bit String = 1 when level_prefix is 0;
Because suffixlength is initialized to 0, there is no suffix for this level;
Suffixlength = 0, so suffixlength ++;
Code output = 0000 100 011 1; 
Encode the next level: level [0] = 3;
Levelcode = 4; level_prefix = 2; Bit String = 001;
Level_suffix = 0; suffixlength = 1; therefore, the bitstream is 0010;
Code output = 0000 100 011 1 0010; 
I = 0, encoding level ends.

5) Number of totalzeros before the last non-zero coefficient of encoding ):
Table 9-7. When totalcoeffs = 5 and total_zero = 3, bit string = 111;
Code output = 0000 100 011 1 0010 111;

6) encode the number (runbefore) of the first zero of each non-zero coefficient:
I = totalcoeffs = 5; zerosleft = total_zeros = 3; Table 9-10:
Encoding in reverse order
Zerosleft = 3, run_before = 1, run_before [4] = 10;
Zerosleft = 2, run_before = 0, run_before [3] = 1;
Zerosleft = 2, run_before = 0, run_before [2] = 1;
Zerosleft = 2, run_before = 1, run_before [1] = 01;
Zerosleft = 1, run_before = 1, run_before [0] the last non-zero coefficient does not need to be expressed by code streams.
Code output = 0000 100 011 1 0010 111 10 1 1 01; 
Encoding is complete. (Cavlc is mainly used for table queries. The standard tables are obtained through a large number of experiments !)

Decoding process:

Receiving code stream: 0000 1000 1110 0101 1110 1101

Computing NC = 1
The detailed decoding process is as follows:
1. Based on the coeff_token and NC Lookup tables (see standard table 9-5), the number of non-zero coefficients totalcoeffs and the number of trailing coefficients trailingones are obtained.
NC = 1 select the corresponding table, coeff_token is 0000100, And the totalcoeffs = 5 trailingones = 3 is displayed in the table.
Output sequence: None
2. parsing the tail Coefficient
The first step shows that there are three drag-and-tail coefficients. Enter the drag-and-tail coefficient symbol to encode the code stream 011. Then, the first and second drag-and-1 coefficients are obtained.
Output:-1,-1, 1 (reverse output)
3. Analyze the magnitude (level) of the non-zero coefficient except the tail Coefficient)
(1) determine the suffix length suffixlength
(2) Get the prefix levelprefix according to the code stream Table 9-6
(3) Obtain

Levelcode = (levelprefix <suffixlength) + levelsuffix
(4) levelcode is an even number level = (LEVEL + 2)/2
Levelcode is an odd number of levels = (-level-1)/2
(5) determine whether to update suffixlegth Based on the set threshold.
Return to the example, in reverse order
I = 0, sufixlefix = 0, the prefix levelprefix of Table 9-6 and 1 = 0, levelcode = 0,

Level = 1, I ++, and sufixle+++ are calculated)
I = 1, sufixlefix = 1, look up table 0010 (3 is prefix, 1 suffix) corresponding to the prefix levelprefix = 2, calculate levelcode = 4, level = 3, I ++
I = 2> = totalcoeffs-trailingones. The non-zero coefficient except the tail coefficient is parsed.
Output: 3, 1,-1,-1
4. parse the number of zeros before each non-zero coefficient
Based on totalcoeffs = 5 and input code stream 111, table 9-7 obtains totalzeros = 3.
Initial I = TotalCoeffs-1 = 4, zeroleft = totalzeros = 3, the number of five non-zero coefficients before the zero Parsing is as follows:
I = 4, zeroleft = 3, according to the code stream 10, table 9-10, runbefor = 1,

Output sequence: 3, 1,-1,-1, 0, 1
I = 3, zeroleft = 3-1 = 2. Based on the code stream 1, check the table runbefore = 0,

Output sequence: 3, 1,-1,-1, 0, 1
I = 2, zeroleft = 2-0 = 2. Based on the code stream 1, check the table runbefore = 0,

Output sequence: 3, 1,-1,-1, 0, 1
I = 1, zeroleft = 2-0 = 2, according to the code stream 01, check the table runbefore = 1,

Output sequence: 3, 0, 1,-1,-1, 0, 1
I = 0, zeroleft = 2-1 = 1, output sequence:,-1,-, 1
5. After decoding, set the remaining elements to 0 and arrange them in reverse order to obtain the 4*4 matrix.
6. Restore to a 4*4 data block.
{
0, 3,-1, 0,
0,-1, 1, 0,
1, 0, 0, 0,
0, 0, 0, 0
}