CN103139558B

CN103139558B - The DC/AC coefficient prediction method of Video Decoder and Video Decoder

Info

Publication number: CN103139558B
Application number: CN201110389023.2A
Authority: CN
Inventors: 付宝玲; 高超; 刘春花
Original assignee: Leadcore Technology Co Ltd
Current assignee: Leadcore Technology Co Ltd
Priority date: 2011-11-30
Filing date: 2011-11-30
Publication date: 2016-08-03
Anticipated expiration: 2031-11-30
Also published as: CN103139558A

Abstract

The present invention provides a kind of DC/AC coefficient prediction method of Video Decoder, and the method has steps of: preset a relief area, relief area internal memory there is a need to prediction block and the most to be predicted piece, pointer PTR sensing the most to be predicted piece, pointer PTR has initial value；Block in relief area is set position number；Initial value is given by the AC coefficient in relief area or DC coefficient；The coefficient of prediction current block, is stored in the numerical value after prediction in relief area；Position number according to block carries out the coefficient prediction of block；Pointer PTR skips to pointer initial value when pointing to end block.Present invention employs cyclic buffer and preserve the macro block data that prediction is used, thus enormously simplify index and the border judgement of predictive coefficient, in the case of memory space is suitable, simplify pre-flow gauge, improve the predictive efficiency in units of block.Present invention also offers a kind of Video Decoder made in aforementioned manners.

Description

The DC/AC coefficient prediction method of Video Decoder and Video Decoder

Technical field

The present invention, about a kind of Video Decoder, particularly relates to a kind of DC/AC coefficient prediction method of Video Decoder and uses the Video Decoder of the method.

Background technology

At present, Techniques for Multimedia Compression has been widely used in various consumer product and communication equipment, such as VCD player, DVD player, mobile phone and MP4 player etc..

The Techniques for Multimedia Compression of a kind of routine is the most up-to-date MPEG compress technique, i.e. MPEG-4.MPEG-4 can provide the audio and video stream service of professional quality in various bandwidth ranges.MPEG-4 is to be defined by Motion Picture Experts Group (MovingPictureExpertsGroup, MPEG), and this specification formally became an international standard in 2000.

Compared with MPEG-1 and MPEG-2, MPEG-4 is more suitable for mutual AV service and remotely monitors, its design object is to make it have wider adaptability and extensibility, the narrowest bandwidth can be utilized to be compressed by frame reconstruction technique and transmit data, it is thus possible to obtain optimal picture quality with minimum data.

MPEG4 is divided into 4 major parts in an encoding process:

1) every frame video is divided into block of pixels；

2) infra-frame prediction, trimming space redundancy；

3) carry out inter prediction by Motion estimation and compensation, cut down time redundancy；

4) by original block and correspondence being predicted the residual block that between block, difference is formed converts, quantifies and entropy code, further trimming space redundancy.

As it is shown in figure 1, macro block P is the least unit of MPEG4 encoding and decoding, every frame comprises multiple macro block P according to resolution difference.Each macro block P can be decomposed into 4 luminance block Bb and 2 chrominance block Bh, and the size of each piece is 8 × 8.The block of one 8 × 8, after dct transform, can be divided into two parts of DC and AC coefficient, and DC coefficient is the low frequency part of image, is most important information, cannot lose.AC coefficient is the HFS of figure, is the description of image detail, and AC coefficient can lose partial information, thus the encoding precision of AC coefficient determines the quality of image and the size of code stream.Macro block P can be divided into frame interior and inter macroblocks by type.Intra-frame macro block uses DC and AC coefficient separately encoded；Inter macroblocks typically uses DC and AC coefficient Unified coding, thus can during the decoding of AC coefficient in the lump by DC coefficient also solution out.

The adaptively selected of the prediction direction of DC and AC coefficient is comparison based on the horizontal and vertical gradient being decoded around block.Fig. 2 shows three blocks being decoded around block.Block " X ", " A ", " B " and " C " represents current block, the block on the left side, the block in the upper left corner and the block of top respectively.If the absolute value of the difference of the DC coefficient of the DC coefficient of block A and block B is less than the absolute value of the DC coefficient of block B and the difference of the DC coefficient of block C, then the DC coefficient of the DC coefficient of block A and block B is the most relatively, in numeric ratio horizontal direction the most in vertical direction closer to, so using the DC coefficient of block C to predict the DC coefficient of current block X；Otherwise, numeric ratio in the horizontal direction is closer to, and i.e. uses the DC coefficient of block A to predict the DC coefficient of current block X.And the prediction of AC coefficient uses the direction identical with the prediction of DC coefficient, the prediction of chrominance block uses the direction identical with the prediction of luminance block.

The Buffer Design of traditional DC/AC coefficient prediction as shown in Figure 3 a and Figure 3 b shows, respectively macro block brightness and the Buffer Design of colourity.Conventional buffer needs to preserve the DC/AC coefficient of all of adjacent 8x8 block of the above macro-block line that current macro is expert at, the DC/AC coefficient of the 8x8 block in the adjacent macro block of previous column of preservation current macro column simultaneously, predict the DC/AC coefficient value of in current macro each piece, it is thus desirable to the memory headroom opened up is: as a example by the QCIF that resolution sizes is 176 × 144, often process a macro block, needing the DC coefficient preserved is (11x2+2+ (11+1) x2)=48, needing the AC coefficient preserved is (48x7)=336, 384 coefficients altogether；But, the retrieving that conventional buffer calculates coefficient used is loaded down with trivial details, the coefficient often calculating a 8x8 block is required to from the beginning of the first address of a upper macro-block line calculates, first macro-block line of the most every frame and the BORDER PROCESSING of first macro block row are loaded down with trivial details, situation in needing framing or outside frame, flow chart branches is more.

The present invention then provides a kind of DC/AC coefficient prediction method of new Video Decoder and Video Decoder in order to improve or to solve above-mentioned problem.

Summary of the invention

The technical problem to be solved in the present invention is to provide a kind of in the case of memory space is suitable, predictive value circulation storage mode and prediction related blocks retrieval mode, thus simplify DC/AC coefficient prediction method and the Video Decoder of the Video Decoder of pre-flow gauge raising predictive efficiency in units of block.

The present invention solves above-mentioned technical problem by such technical scheme:

A kind of DC/AC coefficient prediction method of Video Decoder, the method has steps of:

Using macro block as processing unit, preset six relief areas, the DC coefficient relief area of corresponding brightness block, the AC coefficient relief area of luminance block, the DC coefficient relief area of each chrominance block and the AC coefficient relief area of each chrominance block respectively, having prediction in each relief area needs DC or the AC coefficient of block and needs DC or the AC coefficient of the block predicted；

Initial value is given by the AC coefficient in each relief area or DC coefficient；

DC or the AC coefficient of prediction current block, and the numerical value after prediction is stored in the relief area of each correspondence；

Position according to block carries out the coefficient prediction of block；

The pointer PTR pointing to the block being currently needed for prediction skips to pointer initial value when pointing to end block.

As a modification of the present invention, often one block of decoding, position indicator pointer position moves forward two.

As a modification of the present invention, AC coefficient initial value is 0, and DC initial value is 1024.

As a modification of the present invention, each piece of change situation having predicted rear position indicator pointer PTR is fixed, and uses the mode of form to realize, and implementation of tabling look-up is:

Int16_toffset []={ 1 ,-2,1,2}

PTR+=Offset [n], wherein n=0,1,2,3.

The present invention separately provides a kind of Video Decoder, this Video Decoder uses above-mentioned Forecasting Methodology, this decoder is using macro block as processing unit, and it is provided with six relief areas, the DC coefficient relief area of corresponding brightness block respectively, the AC coefficient relief area of luminance block, the DC coefficient relief area of each chrominance block and the AC coefficient relief area of each chrominance block, , have prediction in each relief area need DC or the AC coefficient of block and need DC or the AC coefficient of the block predicted, position indicator pointer PTR points to the optional position in relief area and has initial value, position indicator pointer PTR skips to pointer initial value when pointing to end block, AC coefficient and DC coefficient in each relief area give initial value.

As a modification of the present invention, often one block of decoding, position indicator pointer PTR position moves forward two.

Int16_toffset []={ 1 ,-2,1,2}

PTR+=Offset [n], wherein n=0,1,2,3.

Compared with prior art, the invention have the advantages that have employed cyclic buffer preserves the macro block data that prediction is used, thus enormously simplify index and the border judgement of predictive coefficient, in the case of memory space is suitable, simplify pre-flow gauge, improve the predictive efficiency in units of block.

Accompanying drawing explanation

Fig. 1 is the schematic diagram of video macro block；

Fig. 2 is to be decoded block and the schematic diagram of three blocks around；

Fig. 3 a and Fig. 3 b is the Buffer Design schematic diagram of traditional DC/AC coefficient prediction；

Fig. 4 be Video Decoder of the present invention DC/AC coefficient prediction method in the schematic diagram data of luminance block relief area before decoded macroblock；

Fig. 5 be Video Decoder of the present invention DC/AC coefficient prediction method in the schematic diagram data of relief area after prediction decoding X1 block；

Fig. 6 be Video Decoder of the present invention DC/AC coefficient prediction method in the schematic diagram data of relief area after prediction decoding X2 block；

Fig. 7 be Video Decoder of the present invention DC/AC coefficient prediction method in the schematic diagram data of relief area after prediction decoding X3 block；

Fig. 8 be Video Decoder of the present invention DC/AC coefficient prediction method in the schematic diagram data of relief area after prediction decoding X4 block；

Fig. 9 be Video Decoder of the present invention DC/AC coefficient prediction method in the distribution schematic diagram of luminance block in macro block；

Figure 10 be Video Decoder of the present invention DC/AC coefficient prediction method in the macro block distribution schematic diagram of simplified mistake.

Detailed description of the invention

Describe the detailed description of the invention of the present invention below in conjunction with the accompanying drawings in detail.

The DC/AC coefficient prediction method of Video Decoder of the present invention is according to rule, the prediction of the DC/AC coefficient of each piece uses 3 blocks adjacent thereto to carry out, in the case of the restriction ratio of internal memory is relatively strict, a two field picture all of DC/AC coefficient can not be preserved, thus only preserve the data that prediction is required.Have employed cyclic buffer for the DC/AC coefficient prediction method of this Video Decoder of the present invention and predict DC and the AC coefficient in units of block.

As a example by the DC coefficient of luminance block, this prediction process is described in detail below.Refer to Fig. 4 to Fig. 8, one block of each expression in figure, with the round dot of hatching represent data not in use by, the round dot of band numeral represents that data are among cyclic buffer, numeral represents this block position number in the buffer, the round dot of white represents that needing future position, X1, X2, X3, X4 is 4 luminance block needing prediction in current macro.Fig. 4 represents the data of the relief area before prediction.

Prediction X1 block needs 4,5,6 three points, and now position represents the position of X1 point surrounding reference block, by the predictor calculation prediction direction of these three point, and carries out the prediction of current 8x8 block according to prediction direction.After X1 has predicted, No. 5 points not in use by, i.e. the upper left point of X1 is not at needs, thus X1 occupies the position of No. 5 points, as it is shown in figure 5, it illustrates the data variation of relief area after prediction X1.

In like manner, it was predicted that X2 block needs 5,6,7 three points.After X2 has predicted, No. 6 points not in use by, thus X2 occupies the position of No. 6 points, and as shown in Figure 6, it illustrates the data variation of relief area after prediction decoding X2.

In like manner, it was predicted that X3 block needs 3,4,5 three points.After X3 has predicted, No. 4 points not in use by, thus X3 occupies the position of No. 4 points, as it is shown in fig. 7, it illustrates the data variation of relief area after prediction X3.

In like manner, it was predicted that X4 block needs 4,5,6 three points.After X4 has predicted, No. 5 points not in use by, thus X4 occupies the position of No. 5 points, and as shown in Figure 8, it illustrates the data variation of relief area after prediction decoding X4.

Comparison diagram 4 to Fig. 8, it can be seen that every time after all luminance block of one macro block of decoding, it will revise 3 points.

In fig. 4, it is assumed that what the current location pointer of relief area pointed to is 5, corresponding point in fig. 8 is 7.Thus it can also be seen that often one macro block of decoding, current location pointer moves forward 2.

The crucial each piece of change situation having predicted rear position indicator pointer (ptr) is fixed, as follows:

X1:ptr+=1

X2:ptr-=2

X3:ptr+=1

X4:ptr+=2

Therefore, by the position of position indicator pointer, when being decoded by macro block order, every time to very convenient and quick with reference to the access of DC value, abandoned the process that the needs in traditional approach recalculate from first address.

Owing to being cyclic buffer, the most down add the when that position indicator pointer ptr pointing to end and will automatically jump to beginning, equally, point to subtract again the when of beginning will automatically jump to end.

The AC coefficient of luminance block is similar with the DC coefficient prediction process of the luminance block above said with the prediction process of the DC/AC coefficient of chrominance block.The first row of vertical prediction reference block or the first row of horizontal forecast reference block is used owing to the AC of each piece predicts, so for corresponding 7 the AC coefficients in position of each DC coefficient in brightness or chrominance AC coefficients relief area.

The DC coefficient Buffer Design of luminance block describes in detail as follows:

1. first initial value is composed in relief area before using cyclic buffer, for AC coefficient, compose initial value 0, initial value 1024 is composed for DC coefficient；

2. when last macro block being decoded to often to go last blocks of, needing continuous 3 numbers below are composed initial values, AC is 0, DC is 1024, simultaneously position indicator pointer ptr+=2 (for inter macroblocks, it is also desirable to similar process)；

3. predict certain value time, it was predicted that after value leave LoopBuffer [ptr] in, A, B, C respective value is:

A=LoopBuffer [ptr-1]

B=LoopBuffer [ptr]

C=LoopBuffer [ptr+1]

LoopBuffer [ptr]=predValue

4. the luminance block number of relief area is:

2 × every row macroblock number+4

5. the initial pointer ptr of relief area can arrange optional position, but must be in the range of relief area；

6. predicted every time after, the change of ptr according to the following rules:

Block1ptr+=1

Block2ptr+=-2

Block3ptr+=1

Block4ptr+=2

Can be in the way of employing to be tabled look-up when implementing, code sample is as follows:

Int16_toffset []={ 1 ,-2,1,2}

Ptr+=offset [n] wherein n=0,1,2,3

For simplifying the analysis, it will be assumed that often row has 4 macro blocks, as it is shown in figure 9, illustrate 9 macro blocks A, B, C, D, E, F, G, H, I, J altogether.

Table 1 is luminance block data situation:

Position	0	1	2	3	4	5	6	7	8	9	10	11	ptr
														Initial value	0	0	0	0	0	0	0	0	0	0	0	0	2
A1	0	0	A1	0	0	0	0	0	0	0	0	0	3
														A2	0	0	A1	A2	0	0	0	0	0	0	0	0	1
A3	0	A3	A1	A2	0	0	0	0	0	0	0	0	2
														A4	0	A3	A4	A2	0	0	0	0	0	0	0	0	4
B1	0	A3	A4	A2	B1	0	0	0	0	0	0	0	5
														B2	0	A3	A4	A2	B1	B2	0	0	0	0	0	0	3
B3	0	A3	A4	B3	B1	B2	0	0	0	0	0	0	4
														B4	0	A3	A4	B3	B4	B2	0	0	0	0	0	0	6
C1	0	A3	A4	B3	B4	B2	C1	0	0	0	0	0	7
														C2	0	A3	A4	B3	B4	B2	C1	C2	0	0	0	0	5
C3	0	A3	A4	B3	B4	C3	C1	C2	0	0	0	0	6
														C4	0	A3	A4	B3	B4	C3	C4	C2	0	0	0	0	8
D1	0	A3	A4	B3	B4	C3	C4	C2	D1	0	0	0	9
														D2	0	A3	A4	B3	B4	C3	C4	C2	D1	D2	0	0	7
D3	0	A3	A4	B3	B4	C3	C4	D3	D1	D2	0	0	8
														D4	0	A3	A4	B3	B4	C3	C4	D3	D4	D2	0	0	10
Ending	0	A3	A4	B3	B4	C3	C4	D3	D4	0	0	0	0
														E1	E1	A3	A4	B3	B4	C3	C4	D3	D4	0	0	0	1
E2	E1	E2	A4	B3	B4	C3	C4	D3	D4	0	0	0	11
														E3	E1	E2	A4	B3	B4	C3	C4	D3	D4	0	0	E3	0
E4	E4	E2	A4	B3	B4	C3	C4	D3	D4	0	0	E3	2
														F1	E4	E2	F1	B3	B4	C3	C4	D3	D4	0	0	E3	3
F2	E4	E2	F1	F2	B4	C3	C4	D3	D4	0	0	E3	1
														F3	E4	F3	F1	F2	B4	C3	C4	D3	D4	0	0	E3	2

F4

E4

F3

F4

F2

B4

C3

C4

D3

D4

0

E3

4

G1

E4

F3

F4

F2

G1

C3

C4

D3

D4

0

E3

5

G2

E4

F3

F4

F2

G1

G2

C4

D3

D4

0

E3

3

G3

E4

F3

F4

G3

G1

G2

C4

D3

D4

0

E3

4

G4

E4

F3

F4

G3

G4

G2

C4

D3

D4

0

E3

6

H1

E4

F3

F4

G3

G4

G2

H1

D3

D4

0

E3

7

H2

E4

F3

F4

G3

G4

G2

H1

H2

D4

0

E3

5

H3

E4

F3

F4

G3

G4

H3

H1

H2

D4

0

E3

6

H4

E4

F3

F4

G3

G4

H3

H4

H2

D4

0

E3

8

Ending

E4

F3

F4

G3

G4

H3

H4

0

E3

10

I1

E4

F3

F4

G3

G4

H3

H4

0

I1

E3

11

I2

E4

F3

F4

G3

G4

H3

H4

0

I1

I2

9

I3

E4

F3

F4

G3

G4

H3

H4

0

I3

I1

I2

10

I4

E4

F3

F4

G3

G4

H3

H4

0

I3

I4

I2

0

J1

F3

F4

G3

G4

H3

H4

0

I3

I4

I2

1

J2

J1

J2

F4

G3

G4

H3

H4

0

I3

I4

I2

11

J3

J1

J2

F4

G3

G4

H3

H4

0

I3

I4

J3

0

J4

J2

F4

G3

G4

H3

H4

0

I3

I4

J3

2

Table 1: luminance block data situation

Chrominance block only one of which 8x8 block, so the prediction relief area of chrominance block is relatively easy, illustrates below as a example by the prediction relief area of the DC coefficient of chrominance block:

2. when last macro block being decoded to often to go last blocks of, needing 1 number below is composed initial value, AC is 0, DC is 1024, pointer ptr+=1 simultaneously

3. predict certain value time, it was predicted that after value leave UVLoopBuffer [ptr] in, A, B, C respective value is:

A=UVLoopBuffer [ptr-1]

B=UVLoopBuffer [ptr]

C=UVLoopBuffer [ptr+1]

UVLoopBuffer [ptr]=predValue

4. relief area chrominance block number:

Often row macroblock number+2

5. the initial pointer Uvptr of relief area can arrange optional position, but must be in the range of relief area.

6. predicted every time after, the change of Uvptr according to the following rules:

Uvptr+=1

For simplifying the analysis, it is assumed that often 4 macro blocks of row, as shown in Figure 10, table 2 expression is the data situation of chrominance block relief area:

Position	0	1	2	3	4	5	UvPtr
								Initial value	0	0	0	0	0	0	1
A	0	A	0	0	0	0	2
								B	0	A	B	0	0	0	3
C	0	A	B	C	0	0	4
								D	0	A	B	C	D	0	5
Ending	0	A	B	C	D	0	0
								E	E	A	B	C	D	0	1
F	E	F	B	C	D	0	2
								G	E	F	G	C	D	0	3
H	E	F	G	H	D	0	4
								Ending	E	F	G	H	0	0	5
I	E	F	G	H	0	I	0
								J	J	F	G	H	0	I	1

Summary and induction is as follows:

(1) as a example by the QCIF that resolution sizes is 176 × 144, often process a macro block, needing the DC coefficient preserved is (11x2+4+ (11+2) x2)=52, needing the AC coefficient preserved is (52x7)=364,416 coefficients altogether, from take data space resource in terms of from the point of view of, the most suitable with traditional approach；

(2) mode of this kind of cyclic buffer makes prediction mode become simple and flexible, and the first indicator index change of interior macroblocks is simple fixing, and hunting zone is concentrated；Border is judged for next and process becomes simple and flexible.

Summary is got up, and the present invention provides a kind of DC/AC coefficient prediction method of Video Decoder, and the method has steps of:

Position according to block carries out the coefficient prediction of block；

The present invention separately provides a kind of Video Decoder using above-mentioned Forecasting Methodology, this decoder is using macro block as processing unit, and it is provided with six relief areas, the DC coefficient relief area of corresponding brightness block respectively, the AC coefficient relief area of luminance block, the DC coefficient relief area of each chrominance block and the AC coefficient relief area of each chrominance block, , have prediction in each relief area need DC or the AC coefficient of block and need DC or the AC coefficient of the block predicted, position indicator pointer PTR points to the optional position in relief area and has initial value, position indicator pointer PTR skips to pointer initial value when pointing to end block, AC coefficient and DC coefficient in each relief area give initial value.

The core concept of the present invention is to preserve DC or the AC coefficient data in the macro block used of prediction according to have employed cyclic buffer, thus enormously simplify the index of predictive coefficient and border judges.

The foregoing is only the better embodiment of the present invention; protection scope of the present invention is not limited with above-mentioned embodiment; in every case those of ordinary skill in the art modify or change according to the equivalence that disclosed content is made, and all should include in the protection domain described in claims.

Claims

1. the DC/AC coefficient prediction method of a Video Decoder, it is characterised in that the method has steps of:

Using macro block as processing unit, preset six relief areas, the DC coefficient relief area of corresponding brightness block, the AC coefficient relief area of luminance block, the DC coefficient relief area of each chrominance block and the AC coefficient relief area of each chrominance block respectively, having prediction in each relief area needs DC or the AC coefficient of block and needs DC or the AC coefficient of the block predicted；Wherein, the described block needing prediction is current block, and described prediction needs block to be the solution code block adjacent with described current block；

DC or the AC coefficient of prediction current block, and the numerical value after prediction is stored in the relief area of each correspondence；Position according to block carries out the coefficient prediction of block；

Pointing to the position indicator pointer PTR being currently needed for predicting block and skip to pointer initial value, and often one block of decoding when pointing to end block, position indicator pointer position moves forward two.

The DC/AC coefficient prediction method of Video Decoder the most according to claim 1, it is characterised in that: AC coefficient initial value is 0, and DC initial value is 1024.

The DC/AC coefficient prediction method of Video Decoder the most according to claim 1, it is characterised in that: each piece of change situation having predicted rear position indicator pointer PTR is fixed, and uses the mode of form to realize, and implementation of tabling look-up is:

Int16_toffset []={ 1 ,-2,1,2}

PTR+=Offset [n], wherein n=0,1,2,3.

4. the Video Decoder of the DC/AC coefficient prediction method using Video Decoder as claimed in claim 1, it is characterized in that, this decoder is using macro block as processing unit, and it is provided with six relief areas, the DC coefficient relief area of corresponding brightness block respectively, the AC coefficient relief area of luminance block, the DC coefficient relief area of each chrominance block and the AC coefficient relief area of each chrominance block, have prediction in each relief area need DC or the AC coefficient of block and need DC or the AC coefficient of the block predicted, position indicator pointer PTR points to the optional position in relief area and has initial value, position indicator pointer PTR skips to pointer initial value when pointing to end block, AC coefficient and DC coefficient in each relief area give initial value, and often one block of decoding, position indicator pointer PTR position moves forward two；Wherein, the described block needing prediction is current block, and described prediction needs block to be the solution code block adjacent with described current block.

Video Decoder the most according to claim 4, it is characterised in that: AC coefficient initial value is 0, and DC initial value is 1024.

Video Decoder the most according to claim 4, it is characterised in that: each piece of change situation having predicted rear position indicator pointer PTR is fixed, and uses the mode of form to realize, and implementation of tabling look-up is:

Int16_toffset []={ 1 ,-2,1,2}

PTR+=Offset [n], wherein n=0,1,2,3.