CN101489136B - Methods and systems for image intra-prediction mode estimation, communication, and organization - Google Patents
Methods and systems for image intra-prediction mode estimation, communication, and organization Download PDFInfo
- Publication number
- CN101489136B CN101489136B CN200910002829.4A CN200910002829A CN101489136B CN 101489136 B CN101489136 B CN 101489136B CN 200910002829 A CN200910002829 A CN 200910002829A CN 101489136 B CN101489136 B CN 101489136B
- Authority
- CN
- China
- Prior art keywords
- predictive mode
- prediction
- mode
- estimation
- pattern
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Lifetime
Links
Abstract
Embodiments of the present invention relate to methods and systems for estimating a pixel prediction mode to be used in a video encoding or decoding process, for communicating pixel prediction modes between video encoders and decoders, and for ordering pixel intra-prediction modes.
Description
The application is dividing an application of following application:
Application number: 03812403.3 (PCT/JP2003/006623)
The applying date: on May 27th, 2003
Denomination of invention: the method and system that image intra-prediction mode is estimated, communicated by letter and organize
Coherent reference
It is 60/319, No. 272 U.S. Patent application that this application requirement is enjoyed the sequence number of submitting on May 28th, 2002, and the sequence number of submitting on July 11st, 2002 is 60/319, No. 390 U.S. Patent application.
Technical field
The specific embodiment of the present invention relates to the intra-prediction of image.
Background technology
Digital video needs a large amount of data for example, in order to show each frame and the whole frame in the digital video sequences (, frame sequence) under non-compress mode.Due to the restriction of bandwidth, the computer network of application transmit non-compressed word video by to(for) great majority is also infeasible.In addition, incompressible digital video needs a large amount of memory spaces.Digital video is encoded according to certain mode conventionally, to reduce the requirement to memory space, and can reduce the demand to bandwidth.
A kind of is interframe (interframe) coding for the technology that digital video is encoded.Interframe encode is utilized, and the different frame in video typically comprises roughly identical with take advantage of the individual pixel choosing of x as x this fact of region.In cataloged procedure, the pixel block in motion vector and a frame interrelates to the motion of the similar pixel block in another frame.Therefore, system just needn't be carried out coding twice to this pixel block, and only need to carry out first encoding, and the motion vector of other pixel blocks of prediction is provided.
Another technology in digital video coding is (intraframe) coding in frame.Intraframe coding is encoded to frame or a part wherein, not with reference to the pixel in other frames.Typical intraframe coding take piece as basis to frame or wherein a part encode.For example, in MEPG-2, the discrete cosine transform of pixel block is used in intraframe coding, then conversion coefficient is encoded.Also there is other intraframe coding technology, for example small echo (wavelet) coding.
Conventionally, these technology are all used sizable tables of data for reference prediction pattern.The internal memory that these tables of data are shared, concerning many machines cheaply, burden is heavy.In addition in treatment facility, for data table memory provides enough internal memories, be also a heavy burden.And, owing to there is large tables of data, finally increased the complexity of system.
Summary of the invention
A first aspect of the present invention provides a kind of system of carrying out the coding of digital video or the inner estimation mode of the estimation pixel that decoding is processed, comprising:
Interior prediction section, along prescribed direction prediction pixel value; And
Estimator, the predictive mode of estimation target of prediction piece,
Described interior prediction section is at least used following predictive mode:
Using the oblique down/left side predictive mode from the level direction that roughly square neck oblique 45 is spent to left down as described prescribed direction;
Using from level roughly to lower right tilt 45 degree directions as described prescribed direction tiltedly/right-hand predictive mode;
Vertical right-hand predictive mode using the direction of 67.5 degree that roughly tilt to lower right from level as described prescribed direction; And
Level below predictive mode using the direction of 22.5 degree that roughly tilt to lower right from level as described prescribed direction,
For described predictive mode according to oblique down/left side predictive mode, tiltedly under/order of right-hand predictive mode, vertical right-hand predictive mode and level below predictive mode carries out pattern numbering with ascending value,
Described estimator by with described target of prediction piece the predictive mode of first of upside adjacency and with the predictive mode of second piece of described target of prediction piece in left side adjacency in pattern number the predictive mode that little predictive mode is estimated as described target of prediction piece,
Described interior prediction section is used the first information and the second information,
When carrying out pixel prediction and equated with the predictive mode of described estimator estimation by the determined in fact best predictive mode of encoder, the described first information represents the predictive mode that uses described estimator to estimate, the information of notifying from code device in order to use the predictive mode of described estimation
When the predictive mode of best predictive mode and described estimator estimation is different actually, described the second information represents to use other the predictive mode outside the predictive mode of described estimation, be in the situation that the predictive mode of described estimation is excluded from mode sequence in essence, performance is the information of the remaining mode sequence of notifying from code device in order to specify the pattern of using in prediction.
A second aspect of the present invention provides a kind of method of estimating the inner estimation mode of pixel, comprising:
Interior prediction steps, along prescribed direction prediction pixel value; And
Estimating step, the predictive mode of estimation target of prediction piece,
Described interior prediction steps is at least used following predictive mode:
Using the oblique down/left side predictive mode from the level direction that roughly square neck oblique 45 is spent to left down as described prescribed direction;
Using from level roughly to lower right tilt 45 degree directions as described prescribed direction tiltedly/right-hand predictive mode;
Vertical right-hand predictive mode using the direction of 67.5 degree that roughly tilt to lower right from level as described prescribed direction; And
Level below predictive mode using the direction of 22.5 degree that roughly tilt to lower right from level as described prescribed direction,
For described predictive mode according to oblique down/left side predictive mode, tiltedly under/order of right-hand predictive mode, vertical right-hand predictive mode and level below predictive mode carries out pattern numbering with ascending value,
Described estimating step by with described target of prediction piece the predictive mode of first of upside adjacency and with the predictive mode of second piece of described target of prediction piece in left side adjacency in pattern number the predictive mode that little predictive mode is estimated as described target of prediction piece,
Described interior prediction steps is used the first information and the second information,
When carrying out pixel prediction and equated with the predictive mode of described estimator estimation by the determined in fact best predictive mode of encoder, the described first information represents the predictive mode that uses described estimator to estimate, the information of notifying from code device in order to use the predictive mode of described estimation
When the predictive mode of best predictive mode and described estimator estimation is different actually, described the second information represents to use other the predictive mode outside the predictive mode of described estimation, be in the situation that the predictive mode of described estimation is excluded from mode sequence in essence, performance is the information of the remaining mode sequence of notifying from code device in order to specify the pattern of using in prediction.
Accompanying drawing explanation
In following accompanying drawing, only describe exemplary embodiment of the present invention, so can not be considered to limit its scope, used accompanying drawing to describe and explain supplementary features of the present invention and details.
Fig. 1 represents the form of some adjacent block.
Fig. 2 represent pixel block and for predict in abutting connection with pixel.
Fig. 3 represents predictive mode direction substantially.
Fig. 4 represents the general direction of the predictive mode in embodiments of the present invention.
Fig. 5 represents the general direction of the predictive mode in embodiments of the present invention.
Fig. 6 represents the general direction of the predictive mode in embodiments of the present invention.
Fig. 7 represents the general direction of the predictive mode in embodiments of the present invention.
Fig. 8 represents the general direction of the predictive mode in embodiments of the present invention.
Fig. 9 means the block diagram of the model estimation in some embodiments of the present invention.
Figure 10 means the block diagram that utilizes in some embodiments the model estimation of the predictive mode set of arranging order.
Figure 11 mean utilize the row be associated with numerical value the block diagram of model estimation of set of order.
Figure 12 means the block diagram that the model estimation in the time cannot obtaining some adjacent block data is selected.
Figure 13 means the block diagram that the mode sequence in some embodiments of the present invention is revised.
Figure 14 means that the pattern of estimation is used to revise the block diagram of the embodiments of the present invention of mode sequence usage.
Figure 15 means that the pattern of estimation is used to revise the block diagram of the embodiments of the present invention of having used the first mode sequence of specific appointment.
Embodiment
Embodiments of the present invention comprise the method and system relevant to the intra-prediction of image.Because all execution modes are all relevant to intra-prediction, term " intra-prediction " and " prediction ", when relating to intra-prediction process, can exchange use.
Embodiments of the present invention are used intraframe coding (intraframe coding) or interior coding (intracoding), to use the spatial redundancy in video image.Because adjacent block has similar attribute conventionally, by reference to the spatial coherence between adjacent block, improved the efficiency of cataloged procedure.The predictive mode using based on adjacent block, target of prediction piece can be used this correlation.
In order more effectively to process, or other reasons, a width digital picture can be divided into a plurality of.As shown in Figure 1, object block " C " 12 may be adjacent to adjacent block " A " 14, and piece " A " 14 is located immediately at the top of object block " C " 12.Another adjacent block " B " 16, is located immediately at the left side of object block " C " 12.With conterminal of object block " C " 12 tools, be also considered to the adjacent block of piece " C " 12.
In different configurations, piece can comprise the pixel of different numbers.For example, piece can comprise 4 * 4 pixel sequences.Piece can also comprise 16 * 16 or 8 * 8 pixel sequences.Other configuration of pixels, comprises square and rectangle sequence, can form piece.
With reference to the related data of pixel in adjacent block, each pixel in can target of prediction piece.This comprises in abutting connection with pixel data or adjacent block data, for predicting those adjacent block or in abutting connection with the predictive mode of pixel.Specific pixel in pixel and object block, can carry out reference with English digital index as shown in Figure 2.Fig. 2 represents 4 * 4 object block, and for example piece " C " 12 comprises 16 pixels, with lowercase character 22, represents.Be located immediately at the pixel in the adjacent block of object block top, use capitalization character 24 to represent.Be located immediately at the pixel in the adjacent block in object block left side, use capitalization character 26 to represent.
Predictive mode comprises instruction or the algorithm for target of prediction piece particular pixels.These patterns can be with reference to one or more adjacent block pixels, described in below pattern description.
Predictive mode
Pattern 0: vertical prediction
A, e, i, m can be predicted by A;
B, f, j, n can be predicted by B;
C, g, k, o can be predicted by C;
D, h, l, p can be predicted by D.
Pattern 1: horizontal forecast
A, b, c, d can be predicted by I;
E, f, g, h can be predicted by J;
I, j, k, l can be predicted by K;
M, n, o, p can be predicted by L.
Pattern 2:DC prediction
If all sampled, A, B, C, D, I, J, K, L can use, and can use all samplings of (A+B+C+D+I+J+K+L+4) > > 3 prediction.If A, B, C, D are unavailable, and I, J, K and L can use, can be with all samplings of (I+J+K+L+2) > > 2 prediction.If I, J, K and L are unavailable, and A, B, C, D can use, can be with all samplings of (A+B+C+D+2) > > 2 prediction.If 8 samplings are all unavailable, the predicted value of all luma samples in piece is 128 so.All the time use this pattern, piece is predicted.
Mode 3: diagonal down/left side prediction
A can be by (A+2B+C+I+2J+K+4) > > 3 predictions;
B, e can be by (B+2C+D+J+2K+L+4) > > 3 predictions;
C, f, i can be by (C+2D+E+K+2L+M+4) > > 3 predictions;
D, g, j, m can be by (D+2E+F+L+2M+N+4) > > 3 predictions;
H, k, n can be by (E+2F+G+M+2N+O+4) > > 3 predictions;
L, o can be by (F+2G+H+N+2O+P+4) > > 3 predictions;
P can be by (G+H+O+P+2) > > 2 predictions;
Pattern 4: under diagonal/right-hand prediction
M can be by (J+2K+L+2) > > 2 predictions;
I, n can be by (I+2J+K+2) > > 2 predictions;
E, j, o can be by (Q+2I+J+2) > > 2 predictions;
A, f, k, p can be by (A+2Q+I+2) > > 2 predictions;
B, g, l can be by (Q+2A+B+2) > > 2 predictions;
C, h can be by (A+2B+C+2) > > 2 predictions;
D can be by (B+2C+D+2) > > 2 predictions;
Pattern 5: vertical left prediction
A, j can be by (Q+A+1) > > 1 prediction;
B, k can be by (A+B+1) > > 1 prediction;
C, l can be by (B+C+1) > > 1 prediction;
D can be by (C+D+1) > > 1 prediction;
E, n can be by (I+2Q+A+2) > > 2 predictions;
F, o can be by (Q+2A+B+2) > > 2 predictions;
G, p can be by (A+2B+C+2) > > 2 predictions;
H can be by (B+2C+D+2) > > 2 predictions;
I can be by (Q+2I+J+2) > > 2 predictions;
M can be by (I+2J+K+2) > > 2 predictions;
Pattern 6: level below prediction
A, g can be by (Q+I+1) > > 1 prediction;
B, h can be by (I+2Q+A+2) > > 2 predictions;
C can be by (Q+2A+B+2) > > 2 predictions;
D can be by (A+2B+C+2) > > 2 predictions;
E, k can be by (I+J+1) > > 1 prediction;
F, l can be by (Q+2I+J+2) > > 2 predictions;
I, o can be by (J+K+1) > > 1 prediction;
J, p can be by (I+2J+K+2) > > 2 predictions;
M can be by (K+L+1) > > 1 prediction;
N can be by (J+2K+L+2) > > 2 predictions;
Mode 7: vertical right-hand prediction
A can be by (2A+2B+J+2K+L+4) > > 3 predictions;
B, i can be by (B+C+1) > > 1 prediction;
C, j can be by (C+D+1) > > 1 prediction;
D, k can be by (D+E+1) > > 1 prediction;
L can be by (E+F+1) > > 1 prediction;
E can be by (A+2B+C+K+2L+M+4) > > 3 predictions;
F, m can be by (B+2C+D+2) > > 2 predictions;
G, n can be by (C+2D+E+2) > > 2 predictions;
H, o can be by (D+2E+F+2) > > 2 predictions;
P can be by (E+2F+G+2) > > 2 predictions;
Pattern 8: level top prediction
A can be by (B+2C+D+2I+2J+4) > > 3 predictions;
B can be by (C+2D+E+I+2J+K+4) > > 3 predictions;
C, e can be by (J+K+1) > > 1 prediction;
D, f can be by (J+2K+L+2) > > 2 predictions;
G, i can be by (K+L+1) > > 1 prediction;
H, j can be by (K+2L+M+2) > > 2 predictions;
L, n can be by (L+2M+N+2) > > 2 predictions;
K, m can be by (L+M+1) > > 1 prediction;
O can be by (M+N+1) > > 1 prediction;
P can be by (M+2N+O+2) > > 2 predictions.
Sequencer procedure, is to take the possibility of every kind of less prediction error of mode producing be basis, and this has improved code efficiency, has reduced the demand to internal memory, and at least can part by mathematical definition.
Each predictive mode can be with each predictive mode of text description for cardinal principle prediction direction in each pattern of above-mentioned title (that is, level top, vertical and diagonal lower left).Predictive mode can use angle direction be described with graphics mode.This angle direction, can describe by the arrow chart outwards radiating from central point, as shown in Figure 3.In this chart, each arrow, and central point all represents predictive mode.The angle that predictive mode is corresponding, has a kind of conventional relation with direction, and this direction is by determining from the direction in abutting connection with pixel mean place to the position of realistic objective pixel through weighting for target of prediction pixel.Yet, in definition and JVT standard above, the description of pattern is had to more precise definition.In Fig. 3, central point 32 represents there is no direction, thus this point can with DC predictive mode associated.Horizontal arrow 34 can represent horizontal forecast pattern.Vertical arrows 36 can represent vertical prediction pattern.From central point, to lower right, extend sideling, approximately become the arrow 38 of 45 degree with horizontal line, represent under diagonal/right-hand (DDR) predictive mode.From central point sideling left below extend, approximately become the arrow 40 of 45 degree with horizontal line, represent diagonal down/left (DDL) predictive mode.DDR and DDL predictive mode, be all called as diagonal predictive mode.
From central point, to upper right, extend sideling, with the arrow 42 of horizontal line into about 22.5 degree, represent level top (HU) predictive mode.From central point, to lower right, extend sideling, with the arrow 44 of horizontal line into about 22.5 degree, represent level below (HD) predictive mode.From central point, to lower right, extend sideling, with the arrow 46 of horizontal line into about 67.5 degree, represent vertical right-hand (VR) predictive mode.From central point direction extension to left down sideling, with the arrow 48 of horizontal line into about 67.5 degree, represent vertical left (VL) predictive mode.Several predictive modes of HU, HD, VR and VL, are called intermediate angle predictive mode jointly.
Can also make many other predictive modes, and use this angle description scheme to be described.
Predictive mode order
The present invention determines, predictive mode should sort according to a kind of mode, and to have reduced the possibility of prediction error consistent with generation conventionally for which.Utilize according to the predictive mode that conventionally produces the order sequence of less prediction error possibility, result data itself has the taxis of more consistent sequence.Further, the communication of pattern can utilize and reduce required internal memory and the coding techniques of bandwidth.For example, the inventor determines horizontal forecast pattern and vertical prediction pattern, with respect to diagonal model, conventionally has more likelihood; And diagonal model has more likelihood than intermediate angle predictive mode.In addition, DC predictive mode (for example, when adjacent block adopts internal schema coding) likelihood is usually less than level and vertical predictive mode, but conventionally than diagonal model, has likelihood.
Piece for noncoherent boundary, for example image border or cutting shape (swipe/swath) border, the order of setting up in some embodiments of the present invention, use conventional term to be described below: vertical and horizontal forecast pattern, compare with DC predictive mode, there is the higher likelihood that produces the prediction error reducing; DC predictive mode is compared with diagonal predictive mode, has the higher likelihood that produces the prediction error reducing; And diagonal predictive mode is compared with intermediate angle predictive mode, there is the higher likelihood that produces the prediction error reducing.
For neighboring edge or border, or disabled of adjacent block or pixel predictive mode data, the order of setting up in some embodiments of the present invention, use conventional term to be described below: DC predictive mode is compared with horizontal forecast pattern with vertical, there is the higher likelihood that produces the prediction error reducing; Vertically compare with diagonal predictive mode with horizontal forecast pattern, there is the higher likelihood that produces the prediction error reducing; Diagonal predictive mode is compared with intermediate angle predictive mode, has the higher likelihood that produces the prediction error reducing.
In the first set of execution mode as shown in Figure 4, can be according to following manner defining mode order:
Pattern 0: vertical prediction
Pattern 1: horizontal forecast
Pattern 2:DC prediction
Mode 3: diagonal down/left side prediction
Pattern 4: under diagonal/right-hand prediction
Pattern 5: level below prediction
Pattern 6: vertical right-hand prediction
Mode 7: vertical left prediction
Pattern 8: level top prediction
In second set of execution mode as shown in Figure 5, can be according to following manner defining mode order:
Pattern 0: horizontal forecast
Pattern 1: vertical prediction
Pattern 2:DC prediction
Mode 3: diagonal down/left side prediction
Pattern 4: under diagonal/right-hand prediction
Pattern 5: level below prediction
Pattern 6: vertical right-hand prediction
Mode 7: vertical left prediction
Pattern 8: level top prediction
In the 3rd set of execution mode as shown in Figure 6, can be according to following manner defining mode order:
Pattern 0: vertical prediction
Pattern 1: horizontal forecast
Pattern 2:DC prediction
Mode 3: diagonal down/left side prediction
Pattern 4: under diagonal/right-hand prediction
Pattern 5: vertical right-hand prediction
Pattern 6: level below prediction
Mode 7: vertical left prediction
Pattern 8: level top prediction
In the 4th set of execution mode as shown in Figure 7, can be according to following manner defining mode order:
Pattern 0: horizontal forecast
Pattern 1: vertical prediction
Pattern 2:DC prediction
Mode 3: diagonal down/left side prediction
Pattern 4: under diagonal/right-hand prediction
Pattern 5: vertical right-hand prediction
Pattern 6: level below prediction
Mode 7: vertical left prediction
Pattern 8: level top prediction
In the 5th set of execution mode as shown in Figure 8, can be according to following manner defining mode order:
Pattern 0:DC prediction
Pattern 1: vertical prediction
Pattern 2: horizontal forecast
Mode 3: diagonal down/left side prediction
Pattern 4: under diagonal/right-hand prediction
Pattern 5: vertical right-hand prediction
Pattern 6: level below prediction
Mode 7: vertical left prediction
Pattern 8: level top prediction
It should be noted that in other execution modes of the present invention, mode sequence can with example in order different.
In some embodiments of the present invention, horizontal forecast (pattern 0) and vertical prediction (pattern 1) may be put upside down if needed.Also will understand, if wished, diagonal down/left side predictive mode and level below predictive mode can be put upside down.In addition, also to understand, if wished, under diagonal/right-hand predictive mode (pattern 5), vertical right-hand predictive mode (pattern 6), vertical left predictive mode (mode 7) and level above predictive mode (pattern 8) can resequence.Further, DC predicts that between the pattern that is placed on 0/ pattern 1 set and 4 set of mode 3/pattern be desirable; If but wish, also can be placed on 4 set of mode 3/pattern, and between pattern 5/ pattern 6/ mode 7/pattern 8 set, or be placed on other position.In addition, if wished, angle mould formula 3-8 can renumber, and on the efficiency of coding, can not make a significant impact.
In embodiments of the present invention, for example, for all (, those use the piece of described prediction scheme), can, to be predicted as basis, respectively predictive mode be resequenced.In addition, be not that all independent pieces (for example those use the piece of described prediction scheme) all will be used prediction benchmark, if for example wished, more than 50%, 75% or 90%.In addition, the order of the predictive mode of different masses can be identical, also can be different.Further, for example, to be predicted as each reorder (, according to a kind of predetermined coherence method) of basic pattern, at least predictive mode 5, pattern 6 and mode 7 are implemented like this, can be adopted any other mode to sort to pattern of surplus.In addition, order the best of predictive mode is, 0,1,2,3,4,5,6,7,8.Can use equally other predefine orders of predictive mode.
In some embodiments of the present invention, can comprise one or multiple tables of data, for enterprise schema's data.These patterns, conventionally according to a kind of sortord tissue, can be used together in each unit of tables of data, so that the set of more sequences to be provided.For example, each entry in tables of data, can comprise digital 0,1,2,3,4,5,6,7,8,9 ordered set.Another kind of selection is that, in each entry of tables of data, in tables of data, the ordered set of numeral comprises 5,6,7,8 or 9 orderly digital collection.For example, the entry of tables of data comprises data acquisition system { 1,2,3,5, the 7} of following entry; { 0,1,2,3,4,5,6}; { 0,1,3,5,6,7,8}, wherein each numeral in set, is all the digital value increasing.For example, another kind of selection, the entry of tables of data comprises data acquisition system { 1,2,3,5, the 7} of following entry; { 0,1,2,3,4,5,6}; 0,1,3,5,6,7,8}, wherein each is gathered at least by 25% of unit, or 35%, or 50%, or 75%, or 90% or included more at high proportion.Like this, with respect to known tables of data method, the anticipation of described table is stronger, and required memory is minimized.
Predefine mode to the sequence of data entry set, should be independent of pixel (for example macro block) in abutting connection with the predictive mode of set.Need be appreciated that in fact tables of data can be " static state ", or while needing basic schema in data, can effectively dynamically generate in whole or in part.Therefore, can use a math equation or algorithm, determine entry, wherein can adopt such technology to create " table ".Therefore, use therein one " tables of data " is not only confined to static table, but further comprises the numerical value set that is used to prediction of determining like this numerical value.
Unfortunately, use new mode digital to replace (for example, in the unit of given data table, numeral being replaced) to the numeral with premode, although may be a kind of improvement, result be still conventionally without the data acquisition system sorting.
Take adjacent block data as basis, estimate pixel predictive mode
With above-mentioned conventionally contrary without the data acquisition system of sequence, even if use, replace, the present invention further draws understanding, should by the highest predictive mode row of likelihood, be first first, secondly by the high predictive mode row of likelihood second, be second, if wished, then according to predefine mode, process all the other patterns.This predefined mode should be independent of the predictive mode in abutting connection with macro block., should be there is the order that the possibility of pattern of surplus reduces (predictive mode that likelihood is the highest, or, if wish the predictive mode that likelihood second is high) in the first-selection sequence to pattern of surplus.
Inner estimation mode based on piece A and piece B, as shown in Figure 1, the inner estimation mode of piece C order can be determined in such a way:
(1) if block A and piece B are " outside " (for example unavailable), only have DC prediction (pattern 2) to allow, so the inner estimation mode of piece C are sequentially { 2}.
(2) if block A is " outside " (for example unavailable), and piece B is not " outside ", for piece C, only has DC prediction (pattern 2) and horizontal forecast (pattern 0) to allow, therefore;
(i) if block B is 2, and the predictive mode of piece C is sequentially { 2,0};
(ii) otherwise, the inner estimation mode of piece C is sequentially { 0,2};
(3) if block A is not " outside ", but piece B is " outside ", for piece C, only has DC prediction (pattern 2) and vertical prediction (pattern 1) to allow, therefore;
(i) if block A is 2, and the inner estimation mode of piece C is sequentially { 2,1};
(ii) otherwise, the inner estimation mode of piece C is sequentially { 1,2};
(4) if block A and piece B are not " outside ",
(i) predictive mode of if block A is less than the predictive mode of piece B, and the inner estimation mode of piece C is sequentially { other patterns are by ascending sort for interior prediction block mode A, interior prediction block mode B } so;
(ii) predictive mode of if block A is more than the predictive mode of piece B, and the inner estimation mode of piece C is sequentially { other patterns are by ascending sort for interior prediction block mode B, interior prediction block mode A } so;
(iii) predictive mode of if block A equals the predictive mode of piece B, and the inner estimation mode of piece C is sequentially { interior prediction block mode A, other patterns are by ascending sort } so.
For example, the predictive mode of if block A is 3, and the predictive mode of piece B is 1, and the inner estimation mode of piece C is sequentially { 1,3,0,2,4,5,6,7,8} so.According to all occurring that likelihood reduces the pattern that (or increasing) arranges, like this auto arrangement of the appearance of pattern of surplus will be still by conventional spread in suitable sequence.According to the sequence permutation from high likelihood to low possibility, can increase forward the likelihood of appropriate prediction.Use entropy coding, can reduce the bit stream that result is encoded.Can also the same arrangement mode that uses other.
Above-mentioned select planning, conceptive be based on following rule, the prediction of if block A is X, and the prediction of piece B is Y, the prediction of piece C may be X or Y so.The prediction of X and/or Y, is positioned at the original position of table, and pattern of surplus is arranged in order thereafter.
State another kind of situation, when the predictive mode of A and B known (comprise this situation, beyond A or B or both described section), while providing the most probable pattern of C, the minimum value of the pattern that piece A and B are used.One in if block A or piece B is " outside ", and most probable pattern equals predictive mode 2.Therefore therefore, being assigned to the sequence of the predictive mode of piece C, is exactly most probable pattern, after follow the pattern of surplus of arranging by ascending order.
With reference to figure 9, embodiments of the present invention have been described.In these execution modes, select an object block to be used for predicting 50.For predicting the predictive mode that first adjacent block is used, determined 52, this piece and the direct adjacency of described object block.For predicting that the predictive mode that second adjacent block used determined 54, this piece also with described object block adjacency.Then, these adjacent block predictive modes are detected 56, higher to determine the minimum likelihood of the prediction error of which generation.
In other execution modes of the present invention, as shown in figure 10, according to producing the still less likelihood of the pattern of prediction error, to predictive mode set sequence 58.Select target piece 60.Be identified for the predictive mode 62 of first adjacent block, and be identified for the predictive mode 64 of second adjacent block.After these two predictive modes, be detected 66, to determine which first occurs in the set of modes of sequence, so just determine corresponding to the high pattern of likelihood that produces less prediction error.
In other execution modes of the present invention, as shown in figure 11, according to the likelihood that produces less prediction error, to predictive mode, set sorts 68.Afterwards, it is associated 70 that the pattern in these ordered set and numerical value carry out, and produces so the higher pattern of likelihood of less prediction error, just with less numerical associations.Afterwards, for predicting that the pattern of first adjacent block determined 72, for predicting that the pattern of second adjacent block also determined 74.Detect afterwards these adjacent block patterns, to determine which pattern and lower numerical associations.This pattern is designated as the evaluation profile of object block prediction.
In other execution mode, as shown in figure 12, the likelihood according to producing less prediction error, sorts to predictive mode set.These patterns in ordered set combine 80 with numerical value, produce so the larger pattern of less prediction error likelihood, combine compared with fractional value with one.Once attempt 82, in order to the pattern of determining that first adjacent block of prediction is used, once attempt 84, to determine second pattern that adjacent block is used of prediction.For example, if predict that predictive mode that first adjacent block used is unavailable 86, just specify an acquiescence predictive mode (DC predictive mode) as the evaluation prediction pattern 90 of described object block.Equally, if unavailable 88 for predicting the predictive mode of second adjacent block, just specify an acquiescence predictive mode (for example DC predictive mode) as the evaluation prediction pattern 90 of described object block.When adjacent block predictive mode is available, can detects these adjacent block patterns, and determine which pattern and a lower numerical associations.After this pattern, be designated as the evaluation profile 92 of the described object block of prediction.
Based on adjacent block data, revise predictive mode order
In some embodiments of the present invention, the adjacent block data that are independent of as above are determined predictive mode sequence, can use adjacent block data modification.With reference to the definite predictive mode of adjacent block data, estimate to be inserted into predictive mode order, in order to this is sequentially modified, to reflect the additional information being obtained from adjacent block data.
In some embodiments, predictive mode estimates take that adjacent block data, as basis, can be inserted directly in predictive mode sequenced collection.Typically, predictive mode is estimated before predictive mode order, to be inserted into or to be considered in advance, and the position of insertion is in the most possible position that produces the pattern that reduces prediction error.Yet in some embodiments, described estimation may be inserted into the diverse location in mode sequence.
In some embodiments of the present invention, as shown in figure 13, a predictive mode is sequentially selected 102, and wherein predictive mode sequential element can produce according to them the order arrangement of the likelihood of less prediction error.In other words, first element in order, represent the less prediction error of the most possible generation of this predictive mode, next element in order, represent that the possibility that this predictive mode produces less prediction error takes second place, etc., until last predictive mode in order, the described predictive mode in this order of representation produces the likelihood minimum of lower prediction error.
As mentioned above, a predictive mode is estimated to be determined 104.Use adjacent block data, determine this estimation.Conventionally, described estimation is the predictive mode using in one or more adjacent block, and this adjacent block is the most possible lower prediction error that produces.Yet, can also adopt other modes to determine prediction.When enough adjacent block predictive mode data are unavailable, for example, at the edge of image or the border of segment, can be in the situation that lack one or more adjacent block or their predictive mode data, the predictive mode of estimating target piece.Under many circumstances, when adjacent block data are limited or disabled time, estimate DC predictive mode.
In some such embodiments, once estimate predictive mode, estimate that predictive mode is just used as the pattern of the lower prediction error of most possible generation, be placed in mode sequence 106.In some embodiments, this is first pattern in order, or the pattern being associated with minimum numerical value.
In other embodiments, estimate that predictive mode can have precedence over preselected mode sequence.In some such embodiments, as shown in figure 14, in encoder, preselected mode sequence designated 110.This sequentially comprises the set of predictive mode, and this set is according to the likelihood that produces lower prediction error, or other are sequentially arranged.Based on adjacent block data, estimate that predictive mode is also by definite 112 for one.In encoder, according to same algorithm or method, determine this estimation predictive mode.Encoder is also determined in fact best predictive mode 114, for based on motion vector or other known technologies, predicts pixel.Then, whether encoder compares predictive mode in fact best described in 116 and estimates predictive mode, identical to determine them.If estimation predictive mode and the in fact predictive mode of the best are same patterns, encoder, to decoder transmitted signal, is estimated predictive mode 118 to use.In some embodiments, estimation predictive mode signal can be carried out with 1 appointment unit, to identify, whether uses estimation model.
If estimate that predictive mode is not in fact best predictive mode, encoder is to decoder transmitted signal, to use other pattern 120.Can carry out by reference to pre-established mode sequence.In encoder deterministic model order, which pattern is to equate most with in fact best predictive mode, and sends to decoder the signal that uses this pattern.
When the predictive mode set of having used through sequence, once obtain further data, the order of set will be resequenced.For example ought determine estimation predictive mode, or when having determined best actual prediction pattern, through the predictive mode set of sequence, will resequence.Like this, modification pattern will be inserted into suddenly the set through sequence, is placed on through before ordered set, or in some cases, the set from through sequence, is deleted.
In some embodiments of the present invention, each pattern in mode sequence, is associated with numerical value in order.In these execution modes, with the numerical value that the pattern of being used is associated, to decoder transmitted signal, to use this predictive mode.In some such embodiments, as shown in figure 15, select to comprise the mode sequence 130 of 9 kinds of predictive modes.The adjacent block data of take are basis, are that the estimation predictive mode of one of 9 kinds of patterns in this order is determined 132.By motion vector method or additive method, determine optimum prediction mode 134.Then, optimum prediction mode compares 136 with estimation predictive mode.If estimate that predictive mode and optimum prediction mode are basic identical, decoder just sends 1 and specifies first signal, to use described estimation predictive mode, specifies first signal identified in decoder kind for this 1.If estimate that predictive mode and optimum prediction mode are unequal, estimation predictive mode is got rid of 140 in essence from order.This eliminating is to resequence by pair set, this estimation model in skip order or realize with additive method.Remaining order comprises 8 patterns effectively, can use one 3 to specify unit's performance.This appointment unit of 3 can be sent to decoder, uses which kind of pattern while predicting to specify.
The term of use in aforementioned detailed description and expression formula are as descriptive but not Limited use, here do not use such term and describe and just get rid of the equivalent terms of feature and the meaning of description or a part wherein illustrating, can confirm that scope of the present invention is only defined and is limited in the scope of following claim.
Claims (2)
1. carry out a system for the coding processing of digital video or the inner estimation mode of the estimation pixel that decoding is processed, comprising:
Interior prediction section, along prescribed direction prediction pixel value; And
Estimator, the predictive mode of estimation target of prediction piece,
Described interior prediction section is at least used following predictive mode:
Using the oblique down/left side predictive mode from the level direction that roughly square neck oblique 45 is spent to left down as described prescribed direction;
Using from level roughly to lower right tilt 45 degree directions as described prescribed direction tiltedly/right-hand predictive mode;
Vertical right-hand predictive mode using the direction of 67.5 degree that roughly tilt to lower right from level as described prescribed direction; And
Level below predictive mode using the direction of 22.5 degree that roughly tilt to lower right from level as described prescribed direction,
For described predictive mode according to oblique down/left side predictive mode, tiltedly under/order of right-hand predictive mode, vertical right-hand predictive mode and level below predictive mode carries out pattern numbering with ascending value,
Described estimator by with described target of prediction piece the predictive mode of first of upside adjacency and with the predictive mode of second piece of described target of prediction piece in left side adjacency in pattern number the predictive mode that little predictive mode is estimated as described target of prediction piece,
Described interior prediction section is used the first information and the second information,
When carrying out pixel prediction and equated with the predictive mode of described estimator estimation by the determined in fact best predictive mode of code device, the described first information represents the predictive mode that uses described estimator to estimate, the information of notifying from described code device in order to use the predictive mode of described estimation
When the predictive mode of best predictive mode and described estimator estimation is different actually, described the second information represents to use other the predictive mode beyond the predictive mode of described estimation, be in the situation that the predictive mode of described estimation is excluded from mode sequence in essence, performance is the information of the remaining mode sequence of notifying from described code device in order to specify the pattern of using in prediction.
2. estimate a method for the inner estimation mode of pixel, comprising:
Interior prediction steps, along prescribed direction prediction pixel value; And
Estimating step, the predictive mode of estimation target of prediction piece,
Described interior prediction steps is at least used following predictive mode:
Using the oblique down/left side predictive mode from the level direction that roughly square neck oblique 45 is spent to left down as described prescribed direction;
Using from level roughly to lower right tilt 45 degree directions as described prescribed direction tiltedly/right-hand predictive mode;
Vertical right-hand predictive mode using the direction of 67.5 degree that roughly tilt to lower right from level as described prescribed direction; And
Level below predictive mode using the direction of 22.5 degree that roughly tilt to lower right from level as described prescribed direction,
For described predictive mode according to oblique down/left side predictive mode, tiltedly under/order of right-hand predictive mode, vertical right-hand predictive mode and level below predictive mode carries out pattern numbering with ascending value,
Described estimating step by with described target of prediction piece the predictive mode of first of upside adjacency and with the predictive mode of second piece of described target of prediction piece in left side adjacency in pattern number the predictive mode that little predictive mode is estimated as described target of prediction piece,
Described interior prediction steps is used the first information and the second information,
When carrying out pixel prediction and equated with the predictive mode of described estimating step estimation by the determined in fact best predictive mode of code device, the described first information represents the predictive mode that uses described estimating step to estimate, the information of notifying from described code device in order to use the predictive mode of described estimation
When the predictive mode of best predictive mode and described estimating step estimation is different actually, described the second information represents to use other the predictive mode outside the predictive mode of described estimation, be in the situation that the predictive mode of described estimation is excluded from mode sequence in essence, performance is the information of the remaining mode sequence of notifying from described code device in order to specify the pattern of using in prediction.
Applications Claiming Priority (15)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US31927202P | 2002-05-28 | 2002-05-28 | |
| US60/319272 | 2002-05-28 | ||
| US60/319,272 | 2002-05-28 | ||
| US31939002P | 2002-07-11 | 2002-07-11 | |
| US60/319,390 | 2002-07-11 | ||
| US60/319390 | 2002-07-11 | ||
| US10/404,298 | 2003-03-31 | ||
| US10/404211 | 2003-03-31 | ||
| US10/404,211 US7289672B2 (en) | 2002-05-28 | 2003-03-31 | Methods and systems for image intra-prediction mode estimation |
| US10/404,298 US7386048B2 (en) | 2002-05-28 | 2003-03-31 | Methods and systems for image intra-prediction mode organization |
| US10/404293 | 2003-03-31 | ||
| US10/404,293 US7236524B2 (en) | 2002-05-28 | 2003-03-31 | Methods and systems for image intra-prediction mode communication |
| US10/404298 | 2003-03-31 | ||
| US10/404,211 | 2003-03-31 | ||
| US10/404,293 | 2003-03-31 |
Related Parent Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB038124033A Division CN100473164C (en) | 2002-05-28 | 2003-05-27 | Methods and systems for image intra-prediction mode estimation, communication, and organization |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101489136A CN101489136A (en) | 2009-07-22 |
| CN101489136B true CN101489136B (en) | 2014-07-16 |
Family
ID=40891776
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN200910002829.4A Expired - Lifetime CN101489136B (en) | 2002-05-28 | 2003-05-27 | Methods and systems for image intra-prediction mode estimation, communication, and organization |
| CN 200910002860 Expired - Lifetime CN101568035B (en) | 2002-05-28 | 2003-05-27 | Methods and systems for image intra-prediction mode estimation, communication and organization |
Family Applications After (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 200910002860 Expired - Lifetime CN101568035B (en) | 2002-05-28 | 2003-05-27 | Methods and systems for image intra-prediction mode estimation, communication and organization |
Country Status (1)
| Country | Link |
|---|---|
| CN (2) | CN101489136B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102300088B (en) * | 2010-06-25 | 2013-11-06 | 财团法人工业技术研究院 | In-frame prediction mode optimization method as well as image compression method and device |
| JP6342116B2 (en) * | 2010-07-15 | 2018-06-13 | シャープ株式会社 | Intra prediction mode estimation device |
| CN113038132B (en) * | 2021-03-18 | 2023-01-20 | 北京奇艺世纪科技有限公司 | Prediction mode determination method and device for Coding Unit (CU) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1198639A (en) * | 1997-03-07 | 1998-11-11 | 次级系统公司 | Intra-macroblock DC and AC coefficient prediction for interlaced digital video |
| WO2000033583A1 (en) * | 1998-11-30 | 2000-06-08 | Microsoft Corporation | Efficient macroblock header coding for video compression |
-
2003
- 2003-05-27 CN CN200910002829.4A patent/CN101489136B/en not_active Expired - Lifetime
- 2003-05-27 CN CN 200910002860 patent/CN101568035B/en not_active Expired - Lifetime
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1198639A (en) * | 1997-03-07 | 1998-11-11 | 次级系统公司 | Intra-macroblock DC and AC coefficient prediction for interlaced digital video |
| WO2000033583A1 (en) * | 1998-11-30 | 2000-06-08 | Microsoft Corporation | Efficient macroblock header coding for video compression |
Non-Patent Citations (3)
| Title |
|---|
| GISLE BJONTEGAARD(EDITOR).H.26L TEST MODEL LONG TERM NUMBER 6 (TML-6)draft0.ITU-T TELECOMMUNICATION STANDARIZATION SECTOR OF ITU,GENEVA,CH,12th meeting,Eibsee,Germany.2001,1-35. * |
| JP特开平5-183758A 1993.07.23 |
| JP特开平8-186823A 1996.07.16 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101568035B (en) | 2012-02-01 |
| CN101489136A (en) | 2009-07-22 |
| CN101568035A (en) | 2009-10-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101087422B (en) | Device for image intra-prediction mode estimation | |
| CN104768011B (en) | Image coding/decoding method and relevant apparatus | |
| CN104811703B (en) | The coding method of video and the coding/decoding method and device of device and video | |
| CN104053010B (en) | Picture coding device and method, and picture decoding apparatus and method | |
| CN101494776B (en) | H.264 code rate control method | |
| CN103931197B (en) | It is determined that the method for the binary code word for conversion coefficient | |
| CN107925759A (en) | Method and apparatus for coding and decoding infra-frame prediction | |
| CN103119935B (en) | Image interpolation method and equipment | |
| CN104935941A (en) | Method for decoding intra prediction mode | |
| CN103931180A (en) | Image decoding apparatus | |
| CN101779463B (en) | Method for processing images and the corresponding electronic device | |
| CN106101729A (en) | Logical frame internal schema name in HEVC Video coding | |
| CN103227923A (en) | Image encoding method and image decoding method | |
| CN102714721A (en) | Method for coding and method for reconstruction of a block of an image | |
| CN102484699B (en) | The method of Code And Decode, the corresponding intrument for Code And Decode are carried out to image | |
| WO2005006256A2 (en) | Method and apparatus for coding motion information | |
| CN101877792A (en) | Intra mode prediction method and device and coder | |
| EP2076047B1 (en) | Video motion estimation | |
| CN103444171A (en) | Methods and devices for encoding and decoding at least one image implementing a selection of pixels to be predicted, and corresponding computer program | |
| CN101489136B (en) | Methods and systems for image intra-prediction mode estimation, communication, and organization | |
| CN103051896A (en) | Mode skipping-based video frequency coding method and mode skipping-based video frequency coding system | |
| CN111970512B (en) | Encoding method, encoding device and device with storage function | |
| CN109618162B (en) | Post-selection prediction method in bandwidth compression | |
| CN116671103A (en) | Intra-frame prediction method, encoder, decoder, and storage medium | |
| KR100780124B1 (en) | Encoding and decoding images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 1132119 Country of ref document: HK |
|
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: GR Ref document number: 1132119 Country of ref document: HK |
|
| CX01 | Expiry of patent term | ||
| CX01 | Expiry of patent term |
Granted publication date: 20140716 |