WO2011124158A1 - 一种简化方向性变换的方法、装置和系统 - Google Patents

一种简化方向性变换的方法、装置和系统 Download PDF

Info

Publication number
WO2011124158A1
WO2011124158A1 PCT/CN2011/072564 CN2011072564W WO2011124158A1 WO 2011124158 A1 WO2011124158 A1 WO 2011124158A1 CN 2011072564 W CN2011072564 W CN 2011072564W WO 2011124158 A1 WO2011124158 A1 WO 2011124158A1
Authority
WO
WIPO (PCT)
Prior art keywords
directional
transformation
geometric
matrix
basic
Prior art date
Application number
PCT/CN2011/072564
Other languages
English (en)
French (fr)
Inventor
杨海涛
区子廉
于浩平
宋锦
Original Assignee
华为技术有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Publication of WO2011124158A1 publication Critical patent/WO2011124158A1/zh

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/103Selection of coding mode or of prediction mode
    • H04N19/11Selection of coding mode or of prediction mode among a plurality of spatial predictive coding modes
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/12Selection from among a plurality of transforms or standards, e.g. selection between discrete cosine transform [DCT] and sub-band transform or selection between H.263 and H.264
    • H04N19/122Selection of transform size, e.g. 8x8 or 2x4x8 DCT; Selection of sub-band transforms of varying structure or type
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/60Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
    • H04N19/61Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding in combination with predictive coding

Definitions

  • Embodiments of the present invention relate to the field of video image compression coding, and in particular, to an image block signal space conversion technique, and specifically, to a method, apparatus, and system for simplifying directional transformation. Background technique
  • DCT discrete cosine transform
  • a predictive mode based directional transform (MDDT) technique is proposed.
  • MDDT uses a directional KL transform to transform the prediction error signal, which can significantly improve the compression efficiency.
  • the technique trains each directional intra prediction error signal of H.264 to obtain a pair of separable KL transform matrices and Ri, and respectively transforms the column vector and the row vector in the ⁇ -sized two-dimensional image block signal X, and finally Obtain a two-dimensional transform coefficient block ⁇ . And Ri are respectively ⁇ matrix.
  • the directional transform of the residual signal can further improve the compression efficiency. This can be achieved by specifying a transformation for each direction. At this time, if a separable KL transform is used for a 4x4 block and eight kinds of directivity transforms are specified, as in the case of intra prediction coding, 16 transform matrices of 4x4 size are required.
  • Compression efficiency is one of the important factors for measuring the coding and decoding efficiency of multimedia technology, and in the case of reducing compression efficiency or maintaining compression efficiency, Some transform codes generate a large number of transform matrixes, and the storage requirements for multimedia codec technology are relatively high. Similarly, the application of a large number of transform matrices in multimedia devices is bound to affect the efficiency of codec. The number of transformation matrices required to be stored can be drastically reduced while maintaining the same compression efficiency.
  • Embodiments of the present invention provide a method and apparatus for directional transform, which reduces the number of transform matrices required for directional transform while maintaining or even improving compression efficiency.
  • the embodiment of the invention provides a directional transform video decoding method, which is characterized in that:
  • Obtaining a target direction of the current block obtaining a corresponding basic direction and one or a set of geometric inverse transforms from a geometric transformation between the basic direction and the target direction and a geometric inverse transformation mapping table;
  • geometrically inverse transforming the block signal to obtain a reconstructed block signal Using the one or a set of geometric inverse transforms, geometrically inverse transforming the block signal to obtain a reconstructed block signal.
  • an embodiment of the present invention further provides a video coding method for directional transformation, which includes: Obtaining a target direction of the current block, obtaining a corresponding basic direction and one or a set of geometric transformations from a geometric transformation and a geometric inverse transformation mapping relationship table between the basic direction and the target direction;
  • the directional transformation matrix is used to directionalally transform the geometrically transformed image block signal to obtain a matrix of transform coefficients.
  • the embodiment of the invention further provides a device for video decoding, which is characterized in that:
  • the storage unit is configured to store a geometric transformation and a geometric inverse transformation mapping relationship table between the basic direction and the target direction; the unit is further configured to store a mapping relationship table between the basic direction and the directional transformation matrix.
  • a directional inverse transform unit configured to obtain a target direction of the current block, obtain a corresponding basic direction and one or a set of geometric inverse transforms from a geometric transformation and a geometric inverse transformation mapping relationship table between the basic direction and the target direction, Obtaining a directional transformation matrix from the mapping relationship table of the basic direction and the directional transformation matrix according to the basic direction, and performing spatial directional inverse transformation on the transform coefficient matrix received by the decoding end by using the directional transformation matrix, Get the block signal.
  • a geometric inverse transform unit configured to perform geometric inverse transform on the block signal by using the one or a set of geometric inverse transforms to obtain a reconstructed block signal.
  • the embodiment of the invention further provides a device for video coding, which is characterized in that:
  • the storage unit is configured to store a geometric transformation and a geometric inverse transformation mapping relationship table between the basic direction and the target direction; the unit is further configured to store a mapping relationship table between the basic direction and the directional transformation matrix.
  • a geometric transformation unit configured to obtain a target direction of the current block, obtain a corresponding basic direction and a geometric transformation or a set of geometric transformations from the geometric transformation and the geometric inverse transformation mapping relationship table between the basic direction and the target direction, One or a set of geometric transformations geometrically transform the image block signal, The geometrically transformed image block signal is obtained.
  • a directional transform unit configured to obtain, according to the basic direction, a corresponding directional transform matrix from a mapping relationship table between the basic direction and a directional transform matrix, and use the directional transform matrix to transform the geometric transformation
  • the image block signal is directional transformed to obtain a matrix of transform coefficients.
  • the embodiment of the present invention further provides a directional translation inter-predictive video decoding method, which is characterized in that:
  • the directional transformation matrix is used to perform directional inversion on the transform coefficient matrix received by the decoding end to obtain a reconstructed block signal.
  • the embodiment of the present invention further provides a directional transformation inter prediction video coding method, which is characterized by including:
  • the image block signal is directionally transformed using the directional transformation matrix to obtain a transform coefficient matrix.
  • the embodiment of the present invention further provides an apparatus for inter-predictive video decoding, which is characterized in that:
  • the storage unit is configured to store a mapping relationship between the target direction and the directional transformation matrix during inter prediction.
  • a directional inverse transform unit configured to use a directional prediction mode between the current frames as a target direction, and obtain a corresponding directional transformation matrix from a mapping relationship table between the target direction and the directional transformation matrix during the inter prediction.
  • the directional transformation matrix performs a directional inverse transformation on the transform coefficient matrix received by the decoding end to obtain a reconstructed block signal.
  • the embodiment of the present invention further provides an apparatus for inter-predictive video coding, which includes:
  • the storage unit is configured to store a mapping relationship between the target direction and the directional transformation matrix during inter prediction.
  • a directional transform unit configured to use a directional prediction mode between the current frame as a target direction, and obtain a corresponding directional transformation matrix from a mapping relationship table between the target direction and the directional transformation matrix during the inter prediction, using the The directional transformation matrix performs directional transformation on the image block signal to obtain a matrix of transform coefficients.
  • FIG. 1 is a schematic diagram of a 4x4 intra prediction mode direction in the prior art H.264 standard.
  • FIG. 2 is a schematic diagram of a prior art inter prediction mode direction
  • FIG. 3 is a schematic diagram of a directional transform design method according to an embodiment of the present invention
  • FIG. 4 is a schematic diagram of a directional transform video coding method according to an embodiment of the present invention
  • FIG. 6 is a schematic diagram of a directional transformation design method for inter prediction according to an embodiment of the present invention
  • Figure ⁇ is a schematic diagram of directional transform video coding for inter prediction according to an embodiment of the present invention.
  • FIG. 8 is a schematic diagram of directional transform video decoding for inter prediction according to an embodiment of the present invention
  • FIG. 9 is a schematic diagram of a directional transform video coding apparatus according to an embodiment of the present invention
  • FIG. 10 is a schematic diagram of a directional transform video decoding apparatus according to an embodiment of the present invention.
  • FIG. 11 is a schematic diagram of a directional transform video coding apparatus for inter prediction according to an embodiment of the present invention.
  • FIG. 12 is a schematic diagram of a directional transform video decoding apparatus for inter prediction according to an embodiment of the present invention.
  • Figure 13 is a comparison of the results of the simulation of the first embodiment and the second embodiment of the present invention with the prior art, wherein (a) shows the performance of the original MDDT; (b) shows the performance of the present invention.
  • Geometric transformation refers to the operation of flipping, transposing, rotating, etc. of the image block signal.
  • geometric inverse transform refers to the geometrically transformed output signal as its input, and restores the operations of flipping, transposing, rotating, etc. required to geometrically transform the original input signal.
  • the geometric (inverse) transform, or geometric (inverse) transform combination is determined by the mapping between the basic direction set and the target direction set, and the statistical properties of the signal need to be considered.
  • Target direction set Refers to the set of all possible directions in the codec process.
  • Basic Direction Set A subset of the target direction set. By geometrically transforming the image block signal having the directional characteristic of the basic direction concentration, an image block signal having the directional characteristic of the target direction concentration can be obtained.
  • Directional transformation matrix A two-dimensional matrix that projects a signal into the geometric space defined by the basis vector within the matrix.
  • the inverse matrix of the directional transformation is the inverse matrix of the directional transformation matrix.
  • the directional transformed signal can be re-projected into the original geometric space.
  • the directional transform matrix can be labeled with the image block signal texture direction.
  • the transform matrix will be a one-dimensional direction along the reference numerals for M e, this time along a direction perpendicular to the reference direction
  • the dimensional transformation matrix can then be correspondingly labeled M 9M .
  • the directional transformation described in the present invention by default, refers to a KL transformation.
  • Geometric transformation and geometric inverse transformation mapping table between the basic direction and the target direction The geometric transformation used to map the direction from the target direction to the direction of the basic direction set, and the direction from the basic direction to the target direction.
  • the inverse geometric transformation used by the direction of the focus It should be emphasized that the table is only one specific implementation manner, and the present invention does not limit other recording methods equivalent to the table.
  • Mapping relationship between the basic direction and the directional transformation matrix The mapping relationship between the basic direction and the directional transformation matrix is recorded.
  • a basic direction corresponds to an inseparable two-dimensional directional transformation matrix;
  • a separable transformation a basic direction corresponds to a one-dimensional column transformation matrix and a one-dimensional row transformation matrix.
  • the block size is 4 x 4 as an example, and other block sizes can be deduced by referring to the theory.
  • the design method of the mapping relationship table used in the simplified directional transformation includes:
  • Step S301 Determine a target direction set, where the target direction set includes all possible direction sets in the coding and decoding process, and different applications exist in different scenarios, such as a 4 ⁇ 4 coding block or an 8 ⁇ 8 coding block.
  • the target direction set can be as shown by the arrow of 0-8 in Fig. 1; when inter prediction is performed, the target direction set can be as shown by the arrow of 0-7 in Fig. 2.
  • Step S302 Determine a basic direction set, and select at least one of the directions from the target direction set as the basic direction set.
  • Step S303 determining a target direction in the target direction set and a base in the basic direction set A geometric transformation mapping relationship between the directions, the geometric transformation includes a horizontal flip transformation, a vertical flip transformation, or a transposition transformation, and the mapping relationship may be recorded in the form of a mapping relationship table, and the mapping relationship table includes The geometric transformation used by the direction in which the target direction is concentrated is mapped to the direction in which the basic direction is concentrated, and the geometric inverse transformation used to map from the direction in which the basic direction is concentrated to the direction in which the target direction is concentrated, and finally obtains basic A geometric transformation and a geometric inverse transformation mapping table between the direction and the target direction.
  • Step S304 Determine a minimum set of the directional transformation matrix according to the direction in which the basic direction is concentrated.
  • Step S305 Determine a mapping relationship between the direction of the basic direction set and the directivity transformation matrix in the minimum set of the directivity transformation matrix, and obtain a mapping table of the basic direction and the directivity transformation matrix.
  • a basic direction corresponds to an inseparable two-dimensional directional transformation matrix
  • a separable transformation corresponds to a one-dimensional column transformation matrix and a one-dimensional row transformation matrix.
  • Step S306 Train the directivity transformation matrix in the minimum set of the directional transformation matrix using the image block samples, and the minimum set of the directional transformation matrix obtained after the training is used for encoding and decoding.
  • the embodiment of the present invention further provides a video coding method for simplifying directional transformation. As shown in FIG. 4, the method includes:
  • Step S401 Obtain a target direction of the current block, select a corresponding basic direction and one or a set of geometric transformations from a geometric transformation and a geometric inverse transformation mapping relationship table between the basic direction and the target direction, and use the one or one Group geometry transforms the image block signal! ⁇ Perform geometric transformation to obtain the image block signal after the geometric transformation;
  • Step S402 Select a corresponding directional transformation matrix from the mapping relationship table between the basic direction and the directional transformation matrix according to the basic direction obtained in step S401, and perform directional transformation using the directional transformation matrix pair.
  • the matrix of transform coefficients ⁇ ⁇ is obtained .
  • entropy coding step of coding For subsequent ⁇ ⁇ 3 ⁇ 4 scanning line corresponding coefficients, entropy coding step of coding.
  • the execution body of the above action can be deployed at the encoding end.
  • the target direction of the current block can be obtained by using a variety of different technologies.
  • the target direction may be the intra prediction direction of the current frame, or may be calculated and acquired by the same method on the codec side;
  • the inter prediction can be indicated in the additional information of the code stream, or can be adaptively calculated at the decoding end. It should be emphasized that the present invention does not limit a particular technique to obtain the target direction of the current block.
  • the embodiment of the present invention further provides a video decoding method for simplifying directional transformation. As shown in FIG. 5, the method includes:
  • Step S501 Obtain a target direction of the current block, select a corresponding basic direction and one or a set of geometric inverse transforms from the geometric transformation and the geometric inverse transformation mapping relationship table between the basic direction and the target direction, according to the basic direction. Selecting a directional transformation matrix from the mapping relationship table between the basic direction and the directional transformation matrix, and performing spatial directional inverse transformation on the transform coefficient matrix ⁇ ⁇ received by the decoding end by using the directional transformation matrix to obtain a block signal ;
  • Step S502 Perform geometric inverse transformation on ⁇ using the one or a set of geometric inverse transforms obtained in step S501 to obtain a reconstructed block signal 3 ⁇ 4.
  • the execution body of the above action can be deployed at the decoding end.
  • the target direction of the current block can be obtained by using a variety of different technologies.
  • the target direction may be the intra prediction direction of the current frame, or may be calculated and acquired by the same method on the codec side;
  • the inter prediction can be indicated in the additional information of the code stream, or can be adaptively calculated at the decoding end. It should be emphasized that the present invention does not limit a particular technique to obtain the target direction of the current block.
  • the embodiment of the present invention uses the 4x4 intra-directional prediction block in H.264 and the separable KL transform used as an example to describe the design and coding and decoding process of the simplified directional transform in detail.
  • Other block sizes can be derived by referring to the theory.
  • FIG. 3 The simplified directional transform design method provided by the embodiment of the present invention is shown in FIG. 3, and the following is a detailed description of each step in the following:
  • Step S301 Determine a target direction set.
  • the target direction set is the eight directions shown in FIG.
  • the numbers in Figure 1 are the intra prediction mode numbers corresponding to each prediction direction.
  • the direction is one-way non-reversible and has a certain orientation.
  • the angle between the mode 4 direction and the reference direction is 45 degrees instead of 1 35 degrees. This is because intra-directional directional prediction is obtained by extrapolation of adjacent reconstructed pixels with a clear orientation.
  • Step S302 Determine a basic direction set.
  • the degree (including the 0 degree angle and the 45 degree angle) is the minimum radians (angle) interval.
  • the target direction is concentrated in all directions by horizontal, vertical flip and/or diagonal transpose operations into the arc (angle) interval, at which point all directions within the interval constitute the basic direction set.
  • the basic direction set is constructed according to the described method, and there is no limit to the number and symmetry of the direction in which the target direction is concentrated.
  • the mode 0 direction is used as the reference direction, and the prediction direction corresponding to the three modes of the basic direction set is 0, 5, and 4 is selected. Modes 0, 5, and 4 correspond to directions of 0 degrees, 30 degrees, and 45 degrees.
  • Step S303 Determine the geometric transformation. Because the symmetry of the image block signal (horizontal and vertical mirror symmetry is mirrored symmetrically with respect to the diagonal direction of the two-dimensional signal) is considered when constructing the new basic direction set, only horizontal flipping, vertical flipping, and transposition are required. Three geometric transformations, you can determine the purpose The mapping relationship between the direction of the target direction and the direction of the basic direction.
  • Table 1 shows the geometric transformations used to map from the direction of the target direction to the direction in the basic direction set, and the inverse of the geometry used to map from the direction of the basic direction to the direction of the target direction set.
  • Table 1 Example of geometric transformation and geometric inverse transformation relationship between basic direction and target direction
  • Step S304 Determine a minimum set of directional transformation matrices.
  • the directional transformation matrix is a one-dimensional directional transformation matrix. Considering the three directions in which the basic directions are concentrated, M is required. , . , M 45 , M 6% M 9Q A total of one-dimensional directional transformation matrix for the image block signal with the direction of the smallest direction to be separable in two directions perpendicular to each other.
  • Step S305 Determine a mapping relationship between a direction of the basic direction set and a directivity transformation matrix in a minimum set of the directivity transformation matrix.
  • two directional transformation matrices are selected for each of the basic directions, and the two directional transformation matrices are a row transformation matrix and a column transformation matrix, respectively.
  • the selection of the two separable one-dimensional directional transformation matrices implemented for the image block is determined by the directionality of the image block signal.
  • a feasible mapping relationship is given in Table 2.
  • Table 2 Example of mapping table between basic direction and directional transformation matrix Basic side corresponding column transformation row transformation
  • Step S 306 The directional transformation matrix in the minimum set of the directional transformation matrix is trained using image block samples, and the minimum set of directional transformation matrices obtained after training is used for encoding and decoding.
  • the process of training the directional transformation matrix includes:
  • C constitutes a one-dimensional directional transformation matrix training sample according to Table 3.
  • B HZV represents all row vector signals or column vector signals of the geometrically transformed image block samples whose prediction direction is X degrees.
  • Step S401 Taking the directional prediction mode in the current frame as the target direction, and selecting the corresponding basic from the geometric transformation and geometric inverse transformation mapping relationship table between the basic direction and the target direction (Table 1 in this embodiment) Direction and one or a set of geometric transformations, using the one or a set of geometric transformations on the image block signal! ⁇ Perform geometric transformation to obtain the geometrically transformed image block signal B s ;
  • Step S402 Select two directional transformation matrices from the mapping relationship table (Table 2 in this embodiment) of the basic direction and the directional transformation matrix according to the basic direction obtained in step S401, the two The directional transformation matrices are a row transformation matrix and a column transformation matrix, respectively, and the directional transformation is performed on the B s using the two directional transformation matrices to obtain a transformation coefficient matrix ⁇ ⁇ .
  • Subsequent coding steps such as corresponding coefficient scanning, entropy coding, etc. are performed.
  • the execution body of the above action can be deployed at the encoding end. Further, the embodiment of the present invention further provides a video decoding method for simplifying directional transformation. As shown in FIG. 5, the following is a detailed description of each step:
  • Step S501 selecting a directional prediction mode in the current frame as a target direction, and selecting a corresponding basic from a geometric transformation and a geometric inverse transformation mapping relationship table (Table 1 in this embodiment) between the basic direction and the target direction.
  • Direction and one or a set of geometric inverse transformations selecting two directional transformation matrices from the mapping relationship table (Table 2 in this embodiment) of the basic direction and the directional transformation matrix according to the basic direction, the two The directional transform matrices are respectively a row transform matrix and a column transform matrix, and the two directional transform matrices are used to perform spatial directional inverse transform on the transform coefficient matrix ⁇ ⁇ received at the decoding end to obtain a block signal.
  • Step S502 performing geometric inverse transformation on the one or a set of geometric inverse transforms obtained in step S501 to obtain a reconstructed block signal ⁇ .
  • the execution body of the above action can be deployed at the decoding end.
  • the number of one-dimensional directional transformation matrices that need to be stored is five, which is significantly smaller than the number of directional transformation matrices that are required to be stored in the prior art.
  • the embodiment of the present invention still uses the 4x4 intra-directional prediction block in H.264 and the separable KL transform used as an example to describe in detail a possible new simplified directional transform design and codec process.
  • Other block sizes may be Derivation by reference theory.
  • FIG. 3 The simplified directional transform design method provided by the embodiment of the present invention is shown in FIG. 3, and the following is a detailed description of each step in the following:
  • Step S301 Determine the target direction set, which is the same as the first step S301.
  • Step S302 Determine a basic direction set. Considering the symmetry of the image signal of the ⁇ ⁇ ( (horizontal and vertical mirror symmetry), you can select the horizontal or vertical direction as the reference direction and select the 0 to 90 degrees from the start of the reference direction (including the 0 degree and 90 degree angles).
  • the interval is the minimum arc (angle) interval.
  • the target direction is concentrated in all directions by horizontal, vertical flip and/or diagonal transposition operations into the arc (angle) interval, at which point all directions within the interval constitute the basic direction set. Selecting a new base direction set reduces the number of geometric transformation operations.
  • the prediction direction corresponding to the five modes of the basic direction set of 0, 5, 4, 6, 1 is selected, that is, 0 degrees, 30 degrees, 45 degrees, 60 degrees and 90 degrees direction.
  • Step S303 Determine a geometric transformation. Since the symmetry of the image block signal (horizontal and vertical mirror symmetry) is considered when constructing the basic direction set, only two geometric transformations, horizontal flip and vertical flip, are required. Because the number of concentrated directions in the basic direction increases, only a few kinds of pre-measurement direction mode image blocks need to be flipped to realize the mapping from the basic direction concentrated direction to the target direction concentrated direction, thereby reducing the number of geometric transformation operations.
  • Step S304 Determine the minimum set of directional transformation matrices. The same as step S304 of the first embodiment.
  • Step S305 In this embodiment, two directional transformation matrices are selected for each of the basic directions, and the two directional transformation matrices are a row transformation matrix and a column transformation matrix, respectively.
  • the choice of the two separable one-dimensional directional transformation matrices implemented for the image block is determined by the directionality of the image block signal. A feasible mapping relationship is given in Table 5.
  • Table 5 Example of mapping table between basic direction and directional transformation matrix
  • Step S306 The directional transformation matrix in the minimum set of the directional transformation matrix is trained using image block samples, and the minimum set of directional transformation matrices obtained after training is used for encoding and decoding.
  • the process of training the directional transformation matrix is the same as that of the first embodiment.
  • a video coding method for simplifying directional transformation is also provided in the embodiment of the present invention. As shown in FIG.
  • Step S401 Taking the directional prediction mode in the current frame as the target direction, selecting a corresponding basic from the geometric transformation and geometric inverse transformation mapping relationship table between the basic direction and the target direction (Table 4 in this embodiment) Direction and one or a set of geometric transformations, geometrically transforming the image block signal B s using the one or a set of geometric transformations to obtain a geometrically transformed image block signal B s ;
  • Step S402 Select two directional transformation matrices from the mapping relationship table (Table 5 in this embodiment) of the basic direction and the directional transformation matrix according to the basic direction obtained in step S401, the two The directional transformation matrices are a row transformation matrix and a column transformation matrix, respectively, and the directional transformation is performed on the B s using the two directional transformation matrices to obtain a transformation coefficient matrix ⁇ ⁇ .
  • entropy coding step of coding For subsequent ⁇ ⁇ 3 ⁇ 4 scanning line corresponding coefficients, entropy coding step of coding.
  • the execution body of the above action can be deployed at the encoding end. Further, the embodiment of the present invention further provides a video decoding method for simplifying directional transformation. As shown in FIG. 5, the following is a detailed description of each step:
  • Step S501 using the directional prediction mode in the current frame as the target direction, and selecting the corresponding basic from the geometric transformation and geometric inverse transformation mapping relationship table between the basic direction and the target direction (Table 4 in this embodiment) Direction and one or a set of geometric inverse transformations, selecting two directional transformation matrices from the mapping relationship table (Table 5 in this embodiment) of the basic direction and the directional transformation matrix according to the basic direction, the two The directional transformation matrices are respectively a row transformation matrix and a column transformation matrix, and the two directional transformation matrices are used to perform spatial directional inverse transformation on the transform coefficient matrix ⁇ received by the decoding end to obtain a block signal.
  • Step S502 Perform geometric inverse transformation on ⁇ using the one or a set of geometric inverse transforms obtained in step S501 to obtain a reconstructed block signal 3 ⁇ 4.
  • the execution body of the above action can be deployed at the decoding end.
  • the number of change matrices is five, which is significantly smaller than the number of directional transform matrices that are required to be stored in the prior art, and the results of simulation tests performed by the first embodiment and the second embodiment of the present invention show that the performance is superior to the prior art.
  • the KL transformation matrix is designed and trained according to the method in the first embodiment and the second embodiment, and the coding and decoding method is integrated into the codec software KTA2. 4rl.
  • the full intra-frame prediction encoding is performed on four WQVGA resolution sequences, and the quantization parameters used are 22, 27, 32, and 37.
  • the simulation results are shown in Fig. 13.
  • the embodiment of the present invention takes the 4x4 intra-directional prediction block and the inseparable KL transform in H.264 as an example, and describes the design and coding and decoding process of the simplified directional transform in detail.
  • Other block sizes can be deduced by referring to the theory.
  • FIG. 3 The simplified directional transform design method provided by the embodiment of the present invention is shown in FIG. 3, and the following is a detailed description of each step in the following:
  • Step S301 Determine the target direction set, which is the same as the first step S301.
  • Step S302 Determine a basic direction set. The same as step S302 of the first embodiment.
  • Step S303 Determine a geometric transformation. The same as the first step S303 of the first embodiment.
  • Step S304 Determine a minimum set of directional transformation matrices.
  • the directional transformation matrix is an inseparable two-dimensional directional transformation matrix. Considering the three directions in the minimum direction set, M 0 , M 3 are required . , M 45 a total of three inseparable two-dimensional directional transformation matrices for spatial transformation of image block signals with the smallest direction of concentration direction.
  • Step S305 In this embodiment, a directional transformation matrix is selected for each of the basic directions, which is a simple one-to-one mapping relationship, as shown in Table 6.
  • Table 6 Example of mapping relationship between basic direction and directional transformation matrix under inseparable transformation Basic square transformation
  • Step S306 The directional transformation matrix in the minimum set of the directional transformation matrix is trained using image block samples, and the minimum set of directional transformation matrices obtained after training is used for encoding and decoding.
  • the process of training the directional transformation matrix includes:
  • each two-dimensional image block signal sample into a one-dimensional vector in a predetermined order, such as a row scan order;
  • Step S401 Taking the directional prediction mode in the current frame as the target direction, and selecting the corresponding basic from the geometric transformation and geometric inverse transformation mapping relationship table between the basic direction and the target direction (Table 1 in this embodiment) a geometric transformation of the image block signal B s using the one or a set of geometric transformations to obtain a geometrically transformed image block signal B s ;
  • Step S402 according to the method obtained in step S401 a basic direction, selecting a two-dimensional directional transformation matrix from the mapping relationship table of the basic direction and the directional transformation matrix (Table 6 in this embodiment), and performing directional transformation using the two-dimensional directional transformation matrix pair , get the transform coefficient matrix ⁇ ⁇ .
  • the execution body of the above action can be deployed at the encoding end.
  • the embodiment of the present invention further provides a video decoding method for simplifying directional transformation. As shown in FIG. 5, the following is a detailed description of each step:
  • Step S501 selecting a directional prediction mode in the current frame as a target direction, and selecting a corresponding basic from a geometric transformation and a geometric inverse transformation mapping relationship table (Table 1 in this embodiment) between the basic direction and the target direction.
  • a geometric transformation and a geometric inverse transformation mapping relationship table (Table 1 in this embodiment) between the basic direction and the target direction.
  • Direction and one or a set of geometric inverse transformations selecting a two-dimensional directional transformation matrix from the mapping relationship table of the basic direction and the directional transformation matrix (Table 6 in this embodiment) according to the basic direction,
  • the two-dimensional directional transformation matrix is subjected to spatial directional inverse transformation on the transform coefficient matrix received by the decoding end to obtain a block signal.
  • Step S502 performing geometric inverse transformation on the one or a set of geometric inverse transforms obtained in step S501 to obtain a reconstructed block signal.
  • the execution body of the above action can be deployed at the decoding end.
  • the number of two-dimensional directional transformation matrices that need to be stored is three, which is significantly smaller than the number of directional transformation matrices that are required to be stored in the prior art.
  • a similar simulation test is performed on the embodiment of the present invention. The simulation test results show that the performance of the embodiment of the present invention is comparable to the prior art.
  • the embodiment of the present invention takes the interframe prediction in H.264 as an example, and considers the case of using the detachable directional KL transform, and describes the design and coding and decoding process of the simplified directional transform in detail.
  • FIG. 3 The simplified directional transform design method provided by the embodiment of the present invention is shown in FIG. 3, and the following is a detailed description of each step in the following:
  • Step S301 Determine a target direction set.
  • the more the number of directions selected by the target direction the more the directional transformation can handle the texture regions with more and finer angle changes. For example, you can choose Select the direction shown in Figure 2 as the target direction set.
  • the numbers in Fig. 2 are the mode numbers corresponding to the respective prediction directions. Modes 0 to 7 correspond to directions of 0, 30, 45, 60, 90, 120, 135, and 150 degrees. It should be noted that in the case of inter prediction, the direction is bidirectionally reversible and does not have a certain orientation.
  • Step S302 Determine a basic direction set. Similar to step S302 in the first embodiment, the basic direction set is selected to be the direction corresponding to the three modes 1, 0, 30 degrees, and 45 degrees.
  • Step S303 Determine a geometric transformation. Similar to step S303 in the first embodiment, a mapping relationship as shown in Table ⁇ is obtained.
  • Table 7 Table of geometric transformations and geometric inverse transformations between basic and target directions during interframe prediction
  • Step S304 Determine a minimum set of directional transformation matrices. It is the same as step S304 in the first embodiment.
  • Step S305 Determine a mapping relationship between a direction of the basic direction set and a directivity transformation matrix in a minimum set of the directivity transformation matrix. Similar to the step S305 in the first embodiment, the mapping relationship as shown in Table 8 is obtained.
  • Table 8 Mapping table of basic direction and directional transformation matrix during interframe prediction Table of basics Corresponding column transformation Row transformation
  • Step S306 The directional transformation matrix in the minimum set of the directional transformation matrix is trained using image block samples, and the minimum set of directional transformation matrices obtained after training is used for encoding and decoding.
  • the process of training the directional transform matrix is the same as step S306 of the first embodiment.
  • a video coding method for simplifying directional transformation is also provided in the embodiment of the present invention. As shown in FIG. 4, the following is a detailed description of each step:
  • Step S401 Select a corresponding basic direction from a geometric transformation and a geometric inverse transformation mapping relationship table (Table 7 in this embodiment) between the basic direction and the target direction with the directional prediction mode between the current frames as the target direction.
  • Direction and one or a set of geometric transformations geometrically transforming the image block signal B s using the one or a set of geometric transformations to obtain a geometrically transformed image block signal B s ;
  • Step S402 Select two directional transformation matrices from the mapping relationship table (Table 8 in this embodiment) of the basic direction and the directional transformation matrix according to the basic direction obtained in step S401, the two The directional transformation matrices are a row transformation matrix and a column transformation matrix, respectively, and the directional transformation is performed using the two directional transformation matrix pairs to obtain a transformation coefficient matrix ⁇ ⁇ .
  • entropy coding step of coding For subsequent ⁇ ⁇ 3 ⁇ 4 scanning line corresponding coefficients, entropy coding step of coding.
  • the execution body of the above action can be deployed at the encoding end. Further, the embodiment of the present invention further provides a video decoding method for simplifying directional transformation. As shown in FIG. 5, the following is a detailed description of each step:
  • Step S501 Select a corresponding basic direction from a geometric transformation and a geometric inverse transformation mapping relationship table (Table 7 in this embodiment) between the basic direction and the target direction with the directional prediction mode between the current frames as the target direction.
  • Direction and one or a set of geometric inverse transformations selecting two from the mapping relationship table of the basic direction and the directional transformation matrix (Table 8 in this embodiment) according to the basic direction a directional transformation matrix, wherein the two directional transformation matrices are a row transformation matrix and a column transformation matrix, respectively, and performing spatial directional inverse transformation on the transform coefficient matrix 0 T received by the decoding end by using the two directional transformation matrices , get the block signal
  • Step S502 performing geometric inverse transformation on the one or a set of geometric inverse transforms obtained in step S501 to obtain a reconstructed block signal.
  • the execution body of the above action can be deployed at the decoding end.
  • the number of one-dimensional directional transformation matrices that need to be stored is five, which is significantly smaller than the number of directional transformation matrices that are required to be stored in the prior art.
  • a similar simulation test is performed on the embodiment of the present invention. The simulation test results show that the performance of the embodiment of the present invention is comparable to the prior art.
  • the embodiment of the present invention takes the interframe prediction in 264 as an example, and considers the case of using the detachable directional KL transform, and describes the design and coding and decoding process of the simplified directional transform in detail.
  • the simplified directional transform design method provided by the embodiment of the present invention includes: Step S601: Determine a target direction set. The same as step S301 of the fourth embodiment.
  • Step S602 Determine a minimum set of the directional transformation matrix. It is the same as step S304 of the fourth embodiment.
  • Step S603 determining a mapping relationship between a direction in which the target direction is concentrated and a directional transformation matrix in a minimum set of the directional transformation matrix, and forming a mapping relationship between a target direction and a directional transformation matrix during inter prediction, such as a table. 9 is shown.
  • the target direction is bidirectionally reversible during interframe prediction, it is possible to select two one-dimensional directions in the direction in which the target direction is concentrated directly according to Table 9.
  • a pair of directional transform matrices the two directional transformation matrices are a row transform matrix and a column transform matrix, respectively, without going through a geometric transform step.
  • Table 9 is an example of a mapping relationship for separable transformations.
  • the target direction corresponds to a two-dimensional inseparable transformation matrix.
  • the table is only one specific implementation manner, and the present invention does not limit other i-recording methods equivalent to the table.
  • Table 9 Example of Mapping Table of Target Direction and Directional Transformation Matrix in Inter Prediction
  • Step S604 Train the directivity transformation matrix in the minimum set of the directional transformation matrix using the image block samples, and the minimum set of the directional transformation matrix obtained after the training is used for encoding and decoding.
  • the process of training the directional transformation matrix is the same as that of step S106 of the fourth embodiment.
  • an inter-predictive video coding method for simplifying directional transformation is also provided in the embodiment of the present invention, as shown in FIG.
  • Step S701 Select a corresponding directional transformation from a mapping relationship table between the target direction and the directional transformation matrix (Table 9 in this embodiment) by using a directional prediction mode between the current frames as a target direction.
  • the matrix, directional transformation of the image block signal B s using the directional transformation matrix, to obtain a matrix of transformation coefficients ⁇ ⁇ .
  • the execution body of the above action can be deployed at the encoding end.
  • the embodiment of the present invention further provides an inter-predictive video decoding method for simplifying directional transformation. As shown in FIG. 8, the method includes:
  • Step S801 Select a corresponding directional transformation from a mapping relationship table between the target direction and the directional transformation matrix (Table 9 in this embodiment) by using a directional prediction mode between the current frames as a target direction.
  • the matrix using the directional transformation matrix, performs a directional inverse transformation on the transform coefficient matrix ⁇ ⁇ received by the decoding end to obtain a reconstructed block signal ⁇ 4. Subsequent prediction compensation, loop filtering and other decoding steps are not changed.
  • the execution body of the above action can be deployed at the decoding end.
  • the number of one-dimensional directional transformation matrices that need to be stored is five, which is significantly smaller than the number of directional transformation matrices that are required to be stored in the prior art.
  • a similar simulation test is performed on the embodiment of the present invention. The simulation test results show that the performance of the embodiment of the present invention is comparable to the prior art.
  • the video encoding apparatus 900 includes:
  • the storage unit 901 the unit stores a geometric transformation and a geometric inverse transformation mapping relationship table between the basic direction and the target direction; the unit further stores a mapping relationship table between the basic direction and the directional transformation matrix; and the geometric transformation unit 902: obtains the current block a target direction, selecting a corresponding basic direction and one or a set of geometric transformations from a geometric transformation and a geometric inverse transformation mapping relationship table between the basic direction and the target direction, using the one or a set of geometric transformations on the image block signal !
  • the video encoding device can be integrated in a circuit or a chip, including a CPU, or a DSP (Digital Signal Processing), or a communication chip.
  • a directional transform unit 903 according to the basic direction obtained in the geometric transformation unit 902, from the mapping relationship between the basic direction and the directional transformation matrix The corresponding directional transformation matrix is selected, and the directional transformation matrix is used to perform directional transformation to obtain a transformation coefficient matrix ⁇ ⁇ .
  • the video encoding device can be integrated in a circuit or a chip, including a CPU, or a DSP (Digital Signal Processing), or a communication chip.
  • the video decoding apparatus 1000 includes:
  • the storage unit 1001 the unit stores a geometric transformation and a geometric inverse transformation mapping relationship table between the basic direction and the target direction; the unit further stores a mapping relationship table between the basic direction and the directional transformation matrix; the directional inverse transformation unit 1002: obtains the current a target direction of the block, selecting a corresponding basic direction and a one or a set of geometric inverse transformations from a geometric transformation and a geometric inverse transformation mapping relationship table between the basic direction and the target direction, according to the basic direction from the basic direction Selecting a directional transformation matrix in the mapping relationship table with the directional transformation matrix, and performing spatial directional inverse transformation on the transformation coefficient matrix received by the decoding end using the directional transformation matrix to obtain a block signal
  • the geometric inverse transform unit 1003 performs the geometric inverse transform using the one or a set of geometric inverse transforms obtained in the directional inverse transform unit 1002 to obtain a reconstructed block signal.
  • the video decoding device can be integrated in a circuit or a chip, including a CPU, or a DSP (Digital Signal Processing, Data Processing, Communication Chip, etc.).
  • a CPU central processing unit
  • a DSP Digital Signal Processing, Data Processing, Communication Chip, etc.
  • the embodiment provides an inter prediction video coding apparatus.
  • the video coding apparatus 1100 includes:
  • the storage unit 1101 the unit stores a mapping relationship table between the target direction and the directional transformation matrix during inter prediction;
  • the directional transform unit 1102 selects a directional transformation matrix from the mapping relationship table between the target direction and the directional transformation matrix in the inter prediction by using the directional prediction mode between the current frames as a target direction, and uses the direction
  • the transform matrix performs a directional transformation on the image block signal 1 ⁇ to obtain a transform coefficient matrix ⁇ ⁇ .
  • the video encoding device can be integrated in a circuit or a chip, including a CPU, or a DSP (Digital Signal Processing), or a communication chip.
  • the video decoding apparatus 1200 includes:
  • the storage unit 1201 is configured to store a mapping relationship between a target direction and a directional transformation matrix during inter prediction
  • the directional inverse transform unit 1002 selects a directional transformation matrix from the mapping relationship table between the target direction and the directional transformation matrix in the inter prediction by using the directional prediction mode between the current frames as a target direction, and uses the The directional transformation matrix performs a directional inverse transformation on the transform coefficient matrix received by the decoding end to obtain a reconstructed block signal ⁇ .
  • the video decoding device can be integrated in a circuit or a chip, including a CPU, or a DSP (Digital Signal Processing, Data Processing, Communication Chip, etc.).
  • a CPU central processing unit
  • a DSP Digital Signal Processing, Data Processing, Communication Chip, etc.
  • the number of directional transformation matrices required to be stored can be significantly reduced.
  • a person skilled in the art can understand that all or part of the process of implementing the above embodiment method can be completed by a computer program to instruct related hardware, and the program can be stored in a computer readable storage medium, the program In execution, the flow of an embodiment of the methods as described above may be included.
  • the storage medium may be a magnetic disk, an optical disk, or a read-only storage memory. Read-Only Memory (ROM) or random access memory (RAM).

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Physics & Mathematics (AREA)
  • Discrete Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)

Description

一种简化方向性变换的方法、 装置和系统 本申请要求了 2010年 4月 9 日提交的, 申请号为 201010147559.9, 发明名称为 "一种简化方向性变换的方法、 装置和系统" 的中国专利申请 的优先权, 其全部内容通过引用结合在本申请中。
技术领域
本发明实施例涉及视频图像压缩编码领域, 尤其涉及一种图像块信号空 间变换技术, 具体为一种简化方向性变换的方法、 装置和系统。 背景技术
现有视频图像压缩编码技术中通常需要对图像块信号 (原始信号或预测 误差信号)进行空间变换, 将信号能量集中于少数几个变换系数中, 以提高 压缩编码效率。 二维可分离离散余弦变换(DCT )被广泛使用, 对图像块信号 进行变换。 若图像块内容较平緩, 或者存在水平或者竖直方向的纹理, DCT 可以显著提高压缩效率; 若图像块中存在其余各种方向性纹理, DCT 压缩效 率显著降低。 为解决这一问题, 已有各种方向性变换技术被提出, 例如可分 离 /不可分离 KL变换, 方向性小波变换等。
结合 H. 264视频编码标准中的方向性帧内预测模式, 一种基于预测模式 的方向性变换(MDDT )技术被提出。 通常, 预测误差信号中依然会存在沿预 测方向的方向性纹理信息。 基于这一特性, MDDT使用方向性 KL变换对预测 误差信号进行变换, 可显著提高压缩效率。 该技术针对 H. 264每一种方向性 帧内预测误差信号训练得到一对可分离 KL变换矩阵 与 Ri,分别对 ΝχΝ大小 的二维图像块信号 X中的列向量与行向量进行变换, 最后得到二维变换系数 块¥。 与 Ri分别为 ΝχΝ矩阵。
Υ = CiXRi1
以 H. 264中 4x4块大小帧内预测为例, 共存在 8种方向性预测模式。 此 时该技术需要存储 8 x 2=16个 4x4大小的变换矩阵。
在帧间预测编码中,对残差信号进行方向性变换可进一步提高压缩效率。 这可通过针对每一种方向规定一种变换实现。 此时, 若对 4x4块使用可分离 KL变换, 并且规定 8种方向性变换, 则与帧内预测编码中的情况相同, 需要 16个 4x4大小的变换矩阵。
在实现本发明的过程中, 发明人发现现有技术中至少存在如下问题: 压缩效率是衡量多媒体技术编解码效率的重要因素之一, 在降低压缩效率 或保持压缩效率不变的情况下, 现有的变换编码产生庞大的变换距阵数 量, 对多媒体编解码技术的存储要求相对较高, 同样, 多媒体设备中数量 繁多的变换矩阵的应用势必影响编解码的效率。 在保持压缩效率不变的情 况下, 可以大幅减少所需存储的变换矩阵数量。
发明内容
本发明实施例提供一种方向性变换的方法和装置, 减少方向性变换所需 的变换矩阵数量, 同时保持甚至提高压缩效率。
本发明实施例提供了一种方向性变换的视频解码方法, 其特征在于, 包 括:
获得当前块的目标方向, 从基本方向和目标方向之间的几何变换和几何 逆变换映射关系表中获得对应的基本方向及一个或一组几何逆变换;
根据所述基本方向从基本方向与方向性变换矩阵的映射关系表中获得方 向性变换矩阵;
使用所述方向性变换矩阵, 对解码端接收到的变换系数矩阵做空间方向 性逆变换, 得到块信号;
使用所述一个或一组几何逆变换, 对所述块信号做几何逆变换, 得到重 建块信号。
另一方面, 本发明实施例还提供一种方向性变换的视频编码方法, 其特 征在于, 包括: 获得当前块的目标方向, 从基本方向和目标方向之间的几何变换和几何 逆变换映射关系表中获得对应的基本方向及一个或一组几何变换;
使用所述一个或一组几何变换对图像块信号进行几何变换, 得到几何变 换后的图像块信号;
根据所述基本方向, 从所述基本方向与方向性变换矩阵的映射关系表中 获得对应的方向性变换矩阵;
使用所述方向性变换矩阵对所述几何变换后的图像块信号做方向性变 换, 得到变换系数矩阵。
再一方面, 本发明实施例还提供一种视频解码的装置, 其特征在于, 包 括:
存储单元, 用于存储基本方向和目标方向之间的几何变换和几何逆变换 映射关系表; 该单元还用于存储基本方向与方向性变换矩阵的映射关系表。
方向性逆变换单元, 用于获得当前块的目标方向, 从所述基本方向和目 标方向之间的几何变换和几何逆变换映射关系表中获得对应的基本方向及一 个或一组几何逆变换, 根据所述基本方向从所述基本方向与方向性变换矩阵 的映射关系表中获得方向性变换矩阵, 使用所述方向性变换矩阵, 对解码端 接收到的变换系数矩阵做空间方向性逆变换, 得到块信号。
几何逆变换单元, 用于使用所述一个或一组几何逆变换, 对所述块信号 做几何逆变换, 得到重建块信号。
再一方面, 本发明实施例还提供一种视频编码的装置, 其特征在于, 包 括:
存储单元, 用于存储基本方向和目标方向之间的几何变换和几何逆变换 映射关系表; 该单元还用于存储基本方向与方向性变换矩阵的映射关系表。
几何变换单元, 用于获得当前块的目标方向, 从所述基本方向和目标方 向之间的几何变换和几何逆变换映射关系表中获得对应的基本方向及一个或 一组几何变换, 使用所述一个或一组几何变换对图像块信号进行几何变换, 得到几何变换后的图像块信号。
方向性变换单元, 用于根据所述基本方向, 从所述基本方向与方向性变 换矩阵的映射关系表中获得对应的方向性变换矩阵, 使用所述方向性变换矩 阵对所述几何变换后的图像块信号做方向性变换, 得到变换系数矩阵。
再一方面, 本发明实施例还提供一种方向性变换的帧间预测视频解码方 法, 其特征在于, 包括:
以当前帧间的方向性预测模式为目标方向, 从帧间预测时目标方向与方 向性变换矩阵的映射关系表中获得对应的方向性变换矩阵;
使用所述方向性变换矩阵对解码端接收到的变换系数矩阵做方向性逆变 换, 得到重建块信号。
再一方面, 本发明实施例还提供一种方向性变换的帧间预测视频编码方 法, 其特征在于, 包括:
以当前帧间的方向性预测模式为目标方向, 从帧间预测时目标方向与方 向性变换矩阵的映射关系表中获得对应的方向性变换矩阵;
使用所述方向性变换矩阵对图像块信号做方向性变换, 得到变换系数矩 阵。
再一方面, 本发明实施例还提供一种帧间预测视频解码的装置, 其特征 在于, 包括:
存储单元, 用于存储帧间预测时目标方向与方向性变换矩阵的映射关系 表。
方向性逆变换单元, 用于以当前帧间的方向性预测模式为目标方向, 从 所述帧间预测时目标方向与方向性变换矩阵的映射关系表中获得对应的方向 性变换矩阵, 使用所述方向性变换矩阵对解码端接收到的变换系数矩阵做方 向性逆变换, 得到重建块信号。
再一方面, 本发明实施例还提供一种帧间预测视频编码的装置, 其特征 在于, 包括: 存储单元, 用于存储帧间预测时目标方向与方向性变换矩阵的映射关系 表。
方向性变换单元, 用于以当前帧间的方向性预测模式为目标方向, 从所 述帧间预测时目标方向与方向性变换矩阵的映射关系表中获得对应的方向性 变换矩阵, 使用所述方向性变换矩阵对图像块信号做方向性变换, 得到变换 系数矩阵。
通过本发明实施例, 使用两个映射关系进行转换, 可以显著减少方向性 变换所需变换矩阵的数量, 同时保持甚至提高压缩编码效率。 附图说明 实施例或现有技术描述中所需要使用的附图作简单地介绍, 显而易见地, 下面描述中的附图仅仅是本发明的一些实施例, 对于本领域普通技术人员 来讲, 在不付出创造性劳动性的前提下, 还可以根据这些附图获得其他的 附图。
图 1是现有技术 H. 264标准中 4x4帧内预测模式方向的示意图
图 2是现有技术帧间预测模式方向的示意图;
图 3是本发明实施例提供的一种方向性变换设计方法的示意图; 图 4是本发明实施例提供的一种方向性变换视频编码方法示意图; 图 5是本发明实施例提供的一种方向性变换视频解码方法示意图; 图 6是本发明实施例提供的一种用于帧间预测的方向性变换设计方法的 示意图;
图 Ί是本发明实施例提供的一种用于帧间预测的方向性变换视频编码示 意图;
图 8是本发明实施例提供的一种用于帧间预测的方向性变换视频解码示 意图; 图 9是本发明实施例提供的一种方向性变换视频编码装置示意图; 图 10是本发明实施例提供的一种方向性变换视频解码装置示意图。
图 11是本发明实施例提供的一种用于帧间预测的方向性变换视频编码 装置示意图。
图 12是本发明实施例提供的一种用于帧间预测的方向性变换视频解码 装置示意图。
图 1 3是本发明实施例一与实施例二仿真的结果与现有技术的对比图,其 中, (a)图表示原有 MDDT的性能; (b)图表示本发明的性能。
具体实施方式
下面将结合本发明实施例中的附图, 对本发明实施例中的技术方案进 行清楚、 完整地描述, 显然, 所描述的实施例仅仅是本发明一部分实施例, 而不是全部的实施例。 基于本发明中的实施例, 本领域普通技术人员在没 有作出创造性劳动前提下所获得的所有其他实施例, 都属于本发明保护的 范围。
为方便描述提出的简化方向性变换装置, 首先解释下面几个概念。
几何变换: 指对图像块信号进行的翻转、 转置、 旋转等操作。 对应地, 几何逆变换指以几何变换输出信号作为其输入, 恢复得到几何变换原输入信 号所需的经翻转、 转置、 旋转等操作。 所采用几何(逆)变换, 或几何(逆) 变换组合则由基本方向集与目标方向集间的映射关系决定, 并且需要考虑信 号统计特性。
目标方向集: 指编解码过程中所有可能的方向的集合。
基本方向集: 为目标方向集的子集。 将具有基本方向集中的方向特性的 图像块信号经几何变换后, 即可得到具有目标方向集中的方向特性的图像块 信号。
方向性变换矩阵: 指 ΝχΝ二维矩阵, 可将信号投影到该矩阵内基向量所 规定的几何空间。 对应的, 方向性变换逆矩阵为方向性变换矩阵的逆矩阵, 可将方向性变换后信号重新投影到原几何空间。 对不可分离变换, 可以使用 图像块信号纹理方向标记方向性变换矩阵。 对于可分离变换需要首先给定基 准方向,若图像块信号纹理方向与基准方向夹角为 θ,则将沿基准方向的一维 变换矩阵标记为 Me, 此时沿与基准方向垂直方向的一维变换矩阵则可对应地 标记为 M9M。 本发明中所述的方向性变换, 缺省地是指 KL变换。
基本方向和目标方向之间的几何变换和几何逆变换映射关系表: 记录了 从目标方向集中的方向映射到基本方向集中的方向所使用的几何变换, 以及 从基本方向集中的方向映射到目标方向集中的方向所使用的几何逆变换。 需 要强调的是, 表只是其中一种具体的实现方式, 本发明不限定其它与表等同 的记录方式。
基本方向与方向性变换矩阵的映射关系表: 记录了基本方向与方向性变 换矩阵的映射关系。 对不可分离变换, 一个基本方向对应着一个不可分离二 维方向性变换矩阵; 对可分离变换, 一个基本方向对应着一个一维的列变换 矩阵和一个一维的行变换矩阵。 需要强调的是, 表只是其中一种具体的实现 方式, 本发明不限定其它与表等同的记录方式。
本发明实施例以块大小采用 4 x 4为例进行阐述,其他块大小可以参照理 论进行推导。
如图 3所示, 本发明实施例提供的简化方向性变换所使用到的映射关系 表的设计方法包括:
步骤 S 301 : 确定目标方向集, 所述的目标方向集包括编解码过程中所有 可能采用的方向集合, 在不同的场景中存在不同的应用, 如对 4 x 4编码块或 者 8 X 8编码块进行帧内预测,其目标方向集可以如图 1中 0 - 8的箭头所示; 当进行帧间预测, 其目标方向集可以如图 2中 0-7的箭头所示。
步骤 S 302 : 确定基本方向集, 从所述目标方向集选择其中至少一个方向 作为基本方向集。
步骤 S 303 : 确定所述目标方向集中的目标方向与所述基本方向集中的基 本方向间的几何变换映射关系, 所述几何变换包括水平翻转变换、 垂直翻转 变换或转置变换等变换方式, 所述映射关系可以采用映射关系表的形式进行 记录, 采用映射关系表则包括从所述目标方向集中的方向映射到所述基本方 向集中的方向所使用的几何变换, 以及从所述基本方向集中的方向映射到所 述目标方向集中的方向所使用的几何逆变换, 最终得到基本方向和目标方向 之间的几何变换和几何逆变换映射关系表。
步骤 S304: 根据所述基本方向集中的方向, 确定方向性变换矩阵的最小 集合。
步骤 S305 : 确定所述基本方向集中的方向与所述方向性变换矩阵的最小 集合中的方向性变换矩阵的映射关系, 得到基本方向与方向性变换矩阵的映 射关系表。 对不可分离变换, 一个基本方向对应着一个不可分离二维方向性 变换矩阵; 对可分离变换, 一个基本方向对应着一个一维的列变换矩阵和一 个一维的行变换矩阵。
步骤 S306: 使用图像块样本训练所述方向性变换矩阵的最小集合中的方 向性变换矩阵, 训练后得到的方向性变换矩阵的最小集合用于编码和解码。 另外, 本发明实施例还提供一种简化方向性变换的视频编码方法, 如图 4所示, 包括:
步骤 S401 : 获得当前块的目标方向, 从所述基本方向和目标方向之间的 几何变换和几何逆变换映射关系表中选择对应的基本方向及一个或一组几何 变换, 使用所述一个或一组几何变换对图像块信号!^进行几何变换, 得到几 何变换后的图像块信号 ;
步骤 S402 : 根据步骤 S401 中获得的所述基本方向, 从所述基本方向与 方向性变换矩阵的映射关系表中选择对应的方向性变换矩阵, 使用所述方向 性变换矩阵对 做方向性变换, 得到变换系数矩阵 Βτ
后续针对 Βτ ¾行相应的系数扫描、 熵编码等编码步骤。 上述动作的执行主体可以部署在编码端。
当前块的目标方向的获取可以采用多种不同的技术, 比如对帧内预测的 情况, 目标方向可以为当前帧的帧内预测方向, 也可以在编解码端采用相同 的方法计算获取; 对帧间预测, 可以在码流的附加信息中指明, 也可以在解 码端自适应地计算获取。 需要强调的是, 本发明不限定某种特定的技术来获 取当前块的目标方向。
进一步地, 本发明实施例还提供一种简化方向性变换的视频解码方法, 如图 5所示, 包括:
步骤 S501 : 获得当前块的目标方向, 从所述基本方向和目标方向之间的 几何变换和几何逆变换映射关系表中选择对应的基本方向及一个或一组几何 逆变换, 根据所述基本方向从所述基本方向与方向性变换矩阵的映射关系表 中选择方向性变换矩阵, 使用所述方向性变换矩阵, 对解码端接收到的变换 系数矩阵 έτ做空间方向性逆变换, 得到块信号 ;
步骤 S502 :使用步骤 S 501中获得的所述一个或一组几何逆变换,对 ^做 几何逆变换, 得到重建块信号 ¾。
后续预测补偿、 环路滤波等解码步骤不做改变。
上述动作的执行主体可以部署在解码端。
当前块的目标方向的获取可以采用多种不同的技术, 比如对帧内预测的 情况, 目标方向可以为当前帧的帧内预测方向, 也可以在编解码端采用相同 的方法计算获取; 对帧间预测, 可以在码流的附加信息中指明, 也可以在解 码端自适应地计算获取。 需要强调的是, 本发明不限定某种特定的技术来获 取当前块的目标方向。
通过本发明实施例, 视频编码和视频解码时, 需要存储的是步骤 S 1 06中 所述训练后得到的方向性变换矩阵的最小集合, 其中的方向性变换矩阵的数 量显著小于现有技术中所需存储的方向性变换矩阵的数量, 同时保持甚至提 高压缩编码效率。 实施例一
本发明实施例以 H. 264中 4x4帧内方向性预测块以及所使用的可分离 KL 变换为例, 详细描述该简化方向性变换的设计与编解码流程, 其他块大小可 以参照理论进行推导。
本发明实施例提供的简化方向性变换设计方法如图 3所示, 下面是对其 中每个步骤的详细说明:
步骤 S 301 : 确定目标方向集。 本实施例中, 目标方向集为图 1所示的 8 个方向。 图 1 中数字为各预测方向所对应的帧内预测模式编号。 需要注意, 在帧内预测的情况下, 所述方向为单向不可反转, 具有确定的指向。 例如, 以模式 0方向作为基准方向, 并沿逆时针方向计算, 模式 4方向与基准方向 的夹角为 45度, 而不是 1 35度。 这是因为帧内方向性预测是通过相邻重建像 素外插得到, 具有明确的指向。
步骤 S 302 : 确定基本方向集。 考虑 Ν χ Ν 图像块信号的对称性(水平、 垂直镜像对称, 与二维信号对角线方向镜像对称) , 可以选择水平或垂直方 向作为基准方向, 并选择基准方向起始的 0度到 45度(包含 0度角与 45度 角) 区间作为最小弧度(角度) 区间。 将目标方向集中所有方向经水平、 垂 直翻转与 (或)对角线转置操作变换到该弧度(角度) 区间内, 此时该区间 内的所有方向组成基本方向集。 需要注意, 按照所描述方法构造基本方向集, 对目标方向集中方向的数量与对称性没有限制。
本实施例中, 以模式 0方向作为基准方向, 选择基本方向集为 0, 5, 4 三个模式所对应的预测方向。 模式 0, 5, 4对应的方向为 0度, 30度与 45 度方向。
步骤 S 303 : 确定几何变换。 因为构造新基本方向集时考虑考虑 Ν χ Ν 图 像块信号的对称性(水平、垂直镜像对称,与二维信号对角线方向镜像对称 ), 因此仅需要水平翻转、 垂直翻转, 以及转置这三种几何变换, 就可以确定目 标方向集中方向与基本方向集中方向的映射关系。
需要注意, 有多个(组) 几何变换可以实现同一种映射。 例如, 从目标 方向 8到基本方向 5可以通过顺时针旋转 90度得到。
一种可行的映射方式在表 1中给出, 其中 T指转置, FH与 Fv分别指以水 平或垂直方向翻转。 表 1 中给出了从目标方向集中的方向映射到基本方向集 中的方向所使用的几何变换, 以及从基本方向集中的方向映射到目标方向集 中的方向所使用的几何逆变换。
表 1: 基本方向和目标方向之间的几何变换和几何逆变换关系表举例
Figure imgf000013_0001
步骤 S 304 : 确定方向性变换矩阵的最小集合。 本实施例中, 方向性变换 矩阵是一维方向性变换矩阵。 考虑基本方向集中的 3个方向, 需要 M。, 。, M45, M6% M9Q共 个一维方向性变换矩阵对具有最小方向集中方向的图像 块信号在互相垂直的两个方向做可分离变换。
步骤 S 305 : 确定所述基本方向集中的方向与所述方向性变换矩阵的最小 集合中的方向性变换矩阵的映射关系。 本实施例中, 为每一个所述基本方向 选择两个方向性变换矩阵, 所述两个方向性变换矩阵分别为行变换矩阵和列 变换矩阵。 对图像块所实施的两个可分离一维方向性变换矩阵的选择则由图 像块信号的方向性决定。 一种可行的映射关系在表 2中给出。
表 2 : 基本方向与方向性变换矩阵的映射关系表举例 基本方 对应 列变换 行变换
向模式 方向 矩阵 矩阵
0 0度 M0 M90
4 45度 M45 M45
5 30度 M30 M60 步骤 S 306 : 使用图像块样本训练所述方向性变换矩阵的最小集合中的方 向性变换矩阵, 训练后得到的方向性变换矩阵的最小集合用于编码和解码。 训练方向性变换矩阵的过程包括:
a. 采集图像块样本;
b. 以当前帧内预测模式所代表的方向作为目标方向, 从表 1中选择一个 (组) 几何变换, 使用所述一个(组) 几何变换对所述图像块样本进行几何 变换;
C 按照表 3构成一维方向性变换矩阵训练样本。 其中 B HZV)表示预测 方向为 X度的所述几何变换后的图像块样本的所有行向量信号或者列向量信 号。
d.对训练样本进行特征值分析或奇异值分解, 即可得到对应一维方向性 变换矩阵。
Figure imgf000014_0001
Figure imgf000014_0002
另外, 本发明实施例还提供的一种简化方向性变换的视频编码方法, 如图 4所示, 下面是对其中每个步骤的详细说明:
步骤 S401 : 以当前帧内的方向性预测模式为目标方向, 从所述基本方向 和目标方向之间的几何变换和几何逆变换映射关系表(本实施例中是表 1 ) 中选择对应的基本方向及一个或一组几何变换, 使用所述一个或一组几何变 换对图像块信号!^进行几何变换, 得到几何变换后的图像块信号 Bs ;
步骤 S402 : 根据步骤 S401 中获得的所述基本方向, 从所述基本方向与 方向性变换矩阵的映射关系表(本实施例中是表 2 ) 中选择两个方向性变换 矩阵, 所述两个方向性变换矩阵分别为行变换矩阵和列变换矩阵, 使用所述 两个方向性变换矩阵对 Bs做方向性变换, 得到变换系数矩阵 Βτ
后续针对 进行相应的系数扫描、 熵编码等编码步骤。
上述动作的执行主体可以部署在编码端。 进一步地, 本发明实施例还提供一种简化方向性变换的视频解码方法, 如图 5所示, 下面是对其中每个步骤的详细说明:
步骤 S501 : 以当前帧内的方向性预测模式为目标方向, 从所述基本方向 和目标方向之间的几何变换和几何逆变换映射关系表(本实施例中是表 1 ) 中选择对应的基本方向及一个或一组几何逆变换, 根据所述基本方向从所述 基本方向与方向性变换矩阵的映射关系表(本实施例中是表 2 ) 中选择两个 方向性变换矩阵,所述两个方向性变换矩阵分别为行变换矩阵和列变换矩阵, 使用所述两个方向性变换矩阵, 对解码端接收到的变换系数矩阵 έτ做空间方 向性逆变换, 得到块信号
步骤 S502 :使用步骤 S501中获得的所述一个或一组几何逆变换,对 做 几何逆变换, 得到重建块信号^。
后续预测补偿、 环路滤波等解码步骤不做改变。
上述动作的执行主体可以部署在解码端。 通过本发明实施例, 视频编码和视频解码时, 需要存储的一维方向性变 换矩阵的数量是 5个, 显著小于现有技术中所需存储的方向性变换矩阵的数 量。 实施例二
本发明实施例仍以 H. 264中 4x4帧内方向性预测块以及所使用的可分离 KL变换为例, 详细描述一种可能的新简化方向性变换的设计与编解码流程, 其他块大小可以参照理论进行推导。
本发明实施例提供的简化方向性变换设计方法如图 3所示, 下面是对其 中每个步骤的详细说明:
步骤 S301 : 确定目标方向集, 与实施例一步骤 S301相同。
步骤 S302 : 确定基本方向集。 考虑 Ν χ Ν 图像块信号的对称性(水平、 垂直镜像对称) , 可以选择水平或垂直方向作为基准方向, 并选择基准方向 起始的 0度到 90度(包含 0度角与 90度角) 区间作为最小弧度(角度) 区 间。 将目标方向集中所有方向经水平、 垂直翻转与 (或)对角线转置操作变 换到该弧度(角度) 区间内, 此时该区间内的所有方向组成基本方向集。 选 取新基本方向集可以减少几何变换操作数量。
在本实施例中, 以模式 0方向作为基准方向, 选择基本方向集为 0, 5, 4, 6, 1五个模式所对应的预测方向, 即 0度, 30度, 45度, 60度与 90度 方向。
步骤 S303: 确定几何变换。 因为构造基本方向集时考虑考虑 Ν χ Ν 图像 块信号的对称性(水平、 垂直镜像对称) , 因此仅需要水平翻转、 垂直翻转 这两种几何变换。 因为基本方向集中方向数量增多, 此时仅需对少数几种预 测方向模式图像块进行翻转操作即可实现从基本方向集中方向与目标方向集 中方向的映射, 从而达到减少几何变换操作数量的目的。
表 4: 基本方向和目标方向之间的几何变换和几何逆变换关系表举例 目标方向 基本方向 几何变换 几何逆变换
0 0 无 无
1 1 无 无
3 4 Fv Fv
4 4 无 无
5 5 无 无
6 6 无 无
7 5 Fv Fv
8 6 FH FH 步骤 S304: 确定方向性变换矩阵的最小集合。 与实施例一步骤 S304相 同。
步骤 S305 : 本实施例中, 为每一个所述基本方向选择两个方向性变换矩 阵, 所述两个方向性变换矩阵分别为行变换矩阵和列变换矩阵。 对图像块所 实施的两个可分离一维方向性变换矩阵的选择则由图像块信号的方向性决 定。 一种可行的映射关系在表 5中给出。
表 5 : 基本方向与方向性变换矩阵的映射关系表举例
Figure imgf000017_0001
步骤 S306: 使用图像块样本训练所述方向性变换矩阵的最小集合中的方 向性变换矩阵, 训练后得到的方向性变换矩阵的最小集合用于编码和解码。 训练方向性变换矩阵的过程与实施例一相同。 另外, 本发明实施例还提供的一种简化方向性变换的视频编码方法, 如图 4所示, 下面是对其中每个步骤的详细说明: 步骤 S401 : 以当前帧内的方向性预测模式为目标方向, 从所述基本方向 和目标方向之间的几何变换和几何逆变换映射关系表(本实施例中是表 4 ) 中选择对应的基本方向及一个或一组几何变换, 使用所述一个或一组几何变 换对图像块信号 Bs进行几何变换, 得到几何变换后的图像块信号 Bs
步骤 S402 : 根据步骤 S401 中获得的所述基本方向, 从所述基本方向与 方向性变换矩阵的映射关系表(本实施例中是表 5 ) 中选择两个方向性变换 矩阵, 所述两个方向性变换矩阵分别为行变换矩阵和列变换矩阵, 使用所述 两个方向性变换矩阵对 Bs做方向性变换, 得到变换系数矩阵 Βτ
后续针对 Βτ ¾行相应的系数扫描、 熵编码等编码步骤。
上述动作的执行主体可以部署在编码端。 进一步地, 本发明实施例还提供一种简化方向性变换的视频解码方法, 如图 5所示, 下面是对其中每个步骤的详细说明:
步骤 S501 : 以当前帧内的方向性预测模式为目标方向, 从所述基本方向 和目标方向之间的几何变换和几何逆变换映射关系表(本实施例中是表 4 ) 中选择对应的基本方向及一个或一组几何逆变换, 根据所述基本方向从所述 基本方向与方向性变换矩阵的映射关系表(本实施例中是表 5 ) 中选择两个 方向性变换矩阵,所述两个方向性变换矩阵分别为行变换矩阵和列变换矩阵, 使用所述两个方向性变换矩阵, 对解码端接收到的变换系数矩阵 τ做空间方 向性逆变换, 得到块信号
步骤 S502 :使用步骤 S 501中获得的所述一个或一组几何逆变换,对 ^做 几何逆变换, 得到重建块信号 ¾。
后续预测补偿、 环路滤波等解码步骤不做改变。
上述动作的执行主体可以部署在解码端。
通过本发明实施例, 视频编码和视频解码时, 需要存储的一维方向性变 换矩阵的数量是 5个, 显著小于现有技术中所需存储的方向性变换矩阵的数 量, 而对本发明实施例一与实施例二进行仿真测试的结果显示其性能优于现 有技术。
按照实施例一与实施例二中的方法设计、训练 KL变换矩阵, 并将所述编 解码方法集成进入编解码软件 KTA2. 4rl。 对 4个 WQVGA分辨率序列进行全帧 内预测编码, 所使用量化参数为 22, 27, 32, and 37 , 仿真结果如图 13所 示。
可以看到, 使用 BD-b i t ra te/BD-PSNR 指标衡量, 本发明的性能 ( -3. 21%/ 0. 24 )甚至略优于原 MDDT性能 ( -3. 00 % /0. 22 ) 。 实施例三
本发明实施例以 H. 264中 4x4帧内方向性预测块以及不可分离 KL变换为 例, 详细描述该简化方向性变换的设计与编解码流程, 其他块大小可以参照 理论进行推导。
本发明实施例提供的简化方向性变换设计方法如图 3所示, 下面是对其 中每个步骤的详细说明:
步骤 S301 : 确定目标方向集, 与实施例一步骤 S301相同。
步骤 S302 : 确定基本方向集。 与实施例一步骤 S302相同。
步骤 S303: 确定几何变换。 与实施例一步骤 S303相同。
步骤 S304: 确定方向性变换矩阵的最小集合。 本实施例中, 方向性变换 矩阵是不可分离二维方向性变换矩阵。 考虑最小方向集中的 3个方向, 需要 M0 , M3。, M45共 3个不可分离二维方向性变换矩阵对具有最小方向集中方向 的图像块信号进行空间变换。
步骤 S305 : 本实施例中, 为每一个所述基本方向选择一个方向性变换矩 阵, 这就是简单的一对一映射关系, 如表 6所示。
表 6: 不可分离变换下基本方向与方向性变换矩阵的映射关系表举例 基本方 对应 变换
向 方向 矩阵
0 0度 M0
4 45度 M45
5 30度 M30 步骤 S306: 使用图像块样本训练所述方向性变换矩阵的最小集合中的方 向性变换矩阵, 训练后得到的方向性变换矩阵的最小集合用于编码和解码。 训练方向性变换矩阵的过程包括:
a. 采集图像块样本;
b. 按照预测模式所代表的方向, 参考表 1对图像块样本进行几何变换; c 按照规定次序例如行扫描次序将每一个二维图像块信号样本组织为 一维向量;
d.对所述信号样本进行特征值分析或奇异值分解, 即可得到对应不可分 离二维方向性变换矩阵。 另外, 本发明实施例还提供的一种简化方向性变换的视频编码方法, 如图 4所示, 下面是对其中每个步骤的详细说明:
步骤 S401 : 以当前帧内的方向性预测模式为目标方向, 从所述基本方向 和目标方向之间的几何变换和几何逆变换映射关系表(本实施例中是表 1 ) 中选择对应的基本方向及一个或一组几何变换, 使用所述一个或一组几何变 换对图像块信号 Bs进行几何变换, 得到几何变换后的图像块信号 Bs; 步骤 S402 : 根据步骤 S401 中获得的所述基本方向, 从所述基本方向与 方向性变换矩阵的映射关系表(本实施例中是表 6 ) 中选择一个二维方向性 变换矩阵, 使用所述二维方向性变换矩阵对 做方向性变换, 得到变换系数 矩阵 Βτ
后续针对 Βτ进行相应的熵编码等编码步骤。 上述动作的执行主体可以部署在编码端。 进一步地, 本发明实施例还提供一种简化方向性变换的视频解码方法, 如图 5所示, 下面是对其中每个步骤的详细说明:
步骤 S501 : 以当前帧内的方向性预测模式为目标方向, 从所述基本方向 和目标方向之间的几何变换和几何逆变换映射关系表(本实施例中是表 1 ) 中选择对应的基本方向及一个或一组几何逆变换, 根据所述基本方向从所述 基本方向与方向性变换矩阵的映射关系表(本实施例中是表 6 ) 中选择一个 二维方向性变换矩阵, 使用所述二维方向性变换矩阵, 对解码端接收到的变 换系数矩阵 ^做空间方向性逆变换, 得到块信号
步骤 S502 :使用步骤 S 501中获得的所述一个或一组几何逆变换,对 做 几何逆变换, 得到重建块信号^。
后续预测补偿、 环路滤波等解码步骤不做改变。
上述动作的执行主体可以部署在解码端。
通过本发明实施例, 视频编码和视频解码时, 需要存储的二维方向性变 换矩阵的数量是 3个, 显著小于现有技术中所需存储的方向性变换矩阵的数 量。 同时, 对本发明实施例进行了类似的仿真测试, 仿真测试结果显示本发 明实施例的性能与现有技术相当。 实施例四
本发明实施例以 H. 264中帧间预测为例,考虑使用可分离方向性 KL变换 的情况, 详细描述该简化方向性变换的设计与编解码流程。
本发明实施例提供的简化方向性变换设计方法如图 3所示, 下面是对其 中每个步骤的详细说明:
步骤 S 301 : 确定目标方向集。 目标方向集中选择的方向数量越多, 则对 应的方向性变换可以处理更多、 更精细角度变化的纹理区域。 例如, 可以选 择如图 2所示方向作为目标方向集。 图 2中数字为各预测方向所对应的模式 编号。 模式 0 ~ 7依次对应方向为 0度、 30度、 45度、 60度、 90度、 120度、 135度、 与 150度。 需要注意, 在帧间预测的情况下, 所述方向为双向可逆, 不具有确定的指向。
步骤 S302: 确定基本方向集。 与实施例一中步骤 S302相似, 选择基本 方向集为 0, 1, 2三个模式所对应方向, 即 0度, 30度与 45度方向。
步骤 S303: 确定几何变换。 与实施例一中步骤 S303相似, 获得如表 Ί 所示的映射关系。
表 7: 帧间预测时基本方向和目标方向之间的几何变换和几何逆变换关系表 举例
Figure imgf000022_0001
步骤 S304: 确定方向性变换矩阵的最小集合。 与实施例一中步骤 S304 相同。
步骤 S305: 确定所述基本方向集中的方向与所述方向性变换矩阵的最小 集合中的方向性变换矩阵的映射关系。 与实施例一中步骤 S305相似, 获得如 表 8所示的映射关系。
表 8: 帧间预测时基本方向与方向性变换矩阵的映射关系表举例 基本方 对应 列变换 行变换
向 方向 矩阵 矩阵
0 0度 M90 M0 1 30度 M60 M30
2 45度 45 45 步骤 S306: 使用图像块样本训练所述方向性变换矩阵的最小集合中的方 向性变换矩阵, 训练后得到的方向性变换矩阵的最小集合用于编码和解码。 训练方向性变换矩阵的过程与实施例一的步骤 S306相同。 另外, 本发明实施例还提供的一种简化方向性变换的视频编码方法, 如图 4所示, 下面是对其中每个步骤的详细说明:
步骤 S401 : 以当前帧间的方向性预测模式为目标方向, 从所述基本方向 和目标方向之间的几何变换和几何逆变换映射关系表(本实施例中是表 7 ) 中选择对应的基本方向及一个或一组几何变换, 使用所述一个或一组几何变 换对图像块信号 Bs进行几何变换, 得到几何变换后的图像块信号 Bs
步骤 S402 : 根据步骤 S401 中获得的所述基本方向, 从所述基本方向与 方向性变换矩阵的映射关系表(本实施例中是表 8 ) 中选择两个方向性变换 矩阵, 所述两个方向性变换矩阵分别为行变换矩阵和列变换矩阵, 使用所述 两个方向性变换矩阵对 做方向性变换, 得到变换系数矩阵 Βτ
后续针对 Βτ ¾行相应的系数扫描、 熵编码等编码步骤。
上述动作的执行主体可以部署在编码端。 进一步地, 本发明实施例还提供一种简化方向性变换的视频解码方法, 如图 5所示, 下面是对其中每个步骤的详细说明:
步骤 S501 : 以当前帧间的方向性预测模式为目标方向, 从所述基本方向 和目标方向之间的几何变换和几何逆变换映射关系表(本实施例中是表 7 ) 中选择对应的基本方向及一个或一组几何逆变换, 根据所述基本方向从所述 基本方向与方向性变换矩阵的映射关系表(本实施例中是表 8 ) 中选择两个 方向性变换矩阵,所述两个方向性变换矩阵分别为行变换矩阵和列变换矩阵, 使用所述两个方向性变换矩阵, 对解码端接收到的变换系数矩阵 0T做空间方 向性逆变换, 得到块信号
步骤 S502 :使用步骤 S 501中获得的所述一个或一组几何逆变换,对 做 几何逆变换, 得到重建块信号^。
后续预测补偿、 环路滤波等解码步骤不做改变。
上述动作的执行主体可以部署在解码端。
通过本发明实施例, 视频编码和视频解码时, 需要存储的一维方向性变 换矩阵的数量是 5个, 显著小于现有技术中所需存储的方向性变换矩阵的数 量。 同时, 对本发明实施例进行了类似的仿真测试, 仿真测试结果显示本发 明实施例的性能与现有技术相当。 实施例五
本发明实施例以 Η. 264中帧间预测为例,考虑使用可分离方向性 KL变换 的情况, 详细描述该简化方向性变换的设计与编解码流程。
需要注意, 因为在帧间预测的情况下, 方向为双向可逆, 不具有确定的 指向。 因此, 在实施例四的基础上可以进行进一步的简化从而构成了本实施 例。
本发明实施例提供的简化方向性变换设计方法如图 6所示, 包括: 步骤 S601 : 确定目标方向集。 与实施例四的步骤 S 301相同。
步骤 S602 : 确定方向性变换矩阵的最小集合。 与实施例四的步骤 S 304 相同。
步骤 S603 : 确定所述目标方向集中的方向与所述方向性变换矩阵的最小 集合中的方向性变换矩阵的映射关系, 构成帧间预测时目标方向与方向性变 换矩阵的映射关系表, 如表 9所示。 如前所述, 由于帧间预测时目标方向为 双向可逆, 因此, 可以按照表 9直接 居目标方向集中的方向选择两个一维 方向性变换矩阵对, 所述两个方向性变换矩阵分别为行变换矩阵和列变换矩 阵, 而不需要经过几何变换步骤。 需要强调的是, 表 9 中是针对可分离变换 的映射关系示例, 对不可分离变换, 目标方向对应的是一个二维不可分离变 换矩阵。 并且, 表只是其中一种具体的实现方式, 本发明不限定其它与表等 同的 i己录方式。
表 9 : 帧间预测时目标方向与方向性变换矩阵的映射关系表举例
Figure imgf000025_0001
步骤 S604: 使用图像块样本训练所述方向性变换矩阵的最小集合中的方 向性变换矩阵, 训练后得到的方向性变换矩阵的最小集合用于编码和解码。 训练方向性变换矩阵的过程与实施例四的步骤 S 106相同。 另外,本发明实施例还提供的一种简化方向性变换的帧间预测视频编码方 法, 如图 Ί所示, 包括:
步骤 S701 : 以当前帧间的方向性预测模式为目标方向, 从所述帧间预测 时目标方向与方向性变换矩阵的映射关系表(本实施例中是表 9 ) 中选择对 应的方向性变换矩阵, 使用所述方向性变换矩阵对图像块信号 Bs做方向性变 换, 得到变换系数矩阵 Βτ
后续针对 Βτ进行相应的系数扫描、 熵编码等编码步骤。
上述动作的执行主体可以部署在编码端。 进一步地, 本发明实施例还提供一种简化方向性变换的帧间预测视频解 码方法, 如图 8所示, 包括:
步骤 S801 : 以当前帧间的方向性预测模式为目标方向, 从所述帧间预测 时目标方向与方向性变换矩阵的映射关系表(本实施例中是表 9 ) 中选择对 应的方向性变换矩阵, 使用所述方向性变换矩阵对解码端接收到的变换系数 矩阵 έτ做方向性逆变换, 得到重建块信号¾。 后续预测补偿、 环路滤波等解码步骤不做改变。
上述动作的执行主体可以部署在解码端。
通过本发明实施例, 视频编码和视频解码时, 需要存储的一维方向性变 换矩阵的数量是 5个, 显著小于现有技术中所需存储的方向性变换矩阵的数 量。 同时, 对本发明实施例进行了类似的仿真测试, 仿真测试结果显示本发 明实施例的性能与现有技术相当。
实施例六
本实施例提供一种视频编码装置, 如图 9所示, 该视频编码装置 900包 括:
存储单元 901 : 该单元存储基本方向和目标方向之间的几何变换和几何 逆变换映射关系表;该单元还存储基本方向与方向性变换矩阵的映射关系表; 几何变换单元 902 : 获得当前块的目标方向, 从所述基本方向和目标方 向之间的几何变换和几何逆变换映射关系表中选择对应的基本方向及一个或 一组几何变换,使用所述一个或一组几何变换对图像块信号!^进行几何变换, 得到几何变换后的图像块信号 Bs; 方向性变换单元 903 : 根据几何变换单元 902中获得的所述基本方向, 从所述基本方向与方向性变换矩阵的映射关系表中选择对应的方向性变换矩 阵, 使用所述方向性变换矩阵对 1¾做方向性变换, 得到变换系数矩阵 Βτ。 该视频编码装置可以集成在电路或芯片中, 包括 CPU、 或 DSP (数字 信号处理, Digital Signal Processing ) 、 或通信芯片等。
通过本发明实施例, 可以显著减少所需存储的方向性变换矩阵的数 量。 实施例七
本实施例提供一种视频解码装置,如图 10所示,该视频解码装置 1000 包括:
存储单元 1001 : 该单元存储基本方向和目标方向之间的几何变换和几何 逆变换映射关系表;该单元还存储基本方向与方向性变换矩阵的映射关系表; 方向性逆变换单元 1002 : 获得当前块的目标方向, 从所述基本方向和目 标方向之间的几何变换和几何逆变换映射关系表中选择对应的基本方向及一 个或一组几何逆变换, 根据所述基本方向从所述基本方向与方向性变换矩阵 的映射关系表中选择方向性变换矩阵, 使用所述方向性变换矩阵, 对解码端 接收到的变换系数矩阵 ^做空间方向性逆变换, 得到块信号
几何逆变换单元 1003: 使用方向性逆变换单元 1002 中获得的所述一个 或一组几何逆变换, 对 做几何逆变换, 得到重建块信号 。
该视频解码装置可以集成在电路或芯片中, 包括 CPU、 或 DSP (数字信号 处理, Di g i ta l S i gna l Proces s ing ) 、 或通信芯片等。
通过本发明实施例, 可以显著减少所需存储的方向性变换矩阵的数量。 实施例八
本实施例提供一种帧间预测视频编码装置, 如图 11所示, 该视频编码装 置 1100包括:
存储单元 1101 : 该单元存储帧间预测时目标方向与方向性变换矩阵的映 射关系表; 方向性变换单元 1102 : 以当前帧间的方向性预测模式为目标方向, 从所 述帧间预测时目标方向与方向性变换矩阵的映射关系表中选择对应的方向性 变换矩阵, 使用所述方向性变换矩阵对图像块信号 1^做方向性变换, 得到变 换系数矩阵 Βτ。 该视频编码装置可以集成在电路或芯片中, 包括 CPU、 或 DSP (数字 信号处理, Digital Signal Processing ) 、 或通信芯片等。
通过本发明实施例, 可以显著减少所需存储的方向性变换矩阵的数 量。 实施例九
本实施例提供一种帧间预测视频解码装置,如图 12所示,该视频解码 装置 1200包括:
存储单元 1201 : 存储帧间预测时目标方向与方向性变换矩阵的映射关系 表;
方向性逆变换单元 1002 : 以当前帧间的方向性预测模式为目标方向, 从 所述帧间预测时目标方向与方向性变换矩阵的映射关系表中选择对应的方向 性变换矩阵, 使用所述方向性变换矩阵对解码端接收到的变换系数矩阵 做 方向性逆变换, 得到重建块信号^。
该视频解码装置可以集成在电路或芯片中, 包括 CPU、 或 DSP (数字信号 处理, Di g i ta l S i gna l Proces s ing ) 、 或通信芯片等。
通过本发明实施例, 可以显著减少所需存储的方向性变换矩阵的数量。 本领域普通技术人员可以理解实现上述实施例方法中的全部或部分 流程, 是可以通过计算机程序来指令相关的硬件来完成, 所述的程序可存 储于一计算机可读取存储介质中, 该程序在执行时, 可包括如上述各方法 的实施例的流程。 其中, 所述的存储介质可为磁碟、 光盘、 只读存储记忆 体 ( Read-Only Memory , ROM ) 或随机存 者 i己忆体 ( Random Access Memory, RAM )等。
领域的普通技术人员应当理解: 其依然可以对本发明实施例的技术方案进行 修改或者等同替换, 而这些修改或者等同替换亦不能使修改后的技术方案脱 离本发明实施例技术方案的精神和范围。

Claims

权 利 要 求 书
1、 一种方向性变换的视频解码方法, 其特征在于, 包括:
获得当前块的目标方向, 从基本方向和目标方向之间的几何变换和几何 逆变换映射关系表中获得对应的基本方向及一个或一组几何逆变换;
根据所述基本方向从基本方向与方向性变换矩阵的映射关系表中获得方 向性变换矩阵;
使用所述方向性变换矩阵, 对解码端接收到的变换系数矩阵做空间方向 性逆变换, 得到块信号;
使用所述一个或一组几何逆变换, 对所述块信号做几何逆变换, 得到重 建块信号。
2、 一种方向性变换的视频编码方法, 其特征在于, 包括:
获得当前块的目标方向, 从基本方向和目标方向之间的几何变换和几何 逆变换映射关系表中获得对应的基本方向及一个或一组几何变换;
使用所述一个或一组几何变换对图像块信号进行几何变换, 得到几何变 换后的图像块信号;
根据所述基本方向, 从所述基本方向与方向性变换矩阵的映射关系表中 获得对应的方向性变换矩阵;
使用所述方向性变换矩阵对所述几何变换后的图像块信号做方向性变 换, 得到变换系数矩阵。
3、 根据权利要求 1或 2所述的方法, 其特征在于, 所述方法还包括: 所述基本方向和目标方向之间的几何变换和几何逆变换映射关系表, 记 录了从目标方向集中的方向映射到基本方向集中的方向所使用的几何变换, 以及从基本方向集中的方向映射到目标方向集中的方向所使用的几何逆变 换。
4、 根据权利要求 3所述的方法, 其特征在于, 所述方法还包括: 所述几何变换包括水平翻转、 垂直翻转、 转置这三种几何变换。
5、 根据权利要求 3所述的方法, 其特征在于, 所述方法还包括: 所述基本方向集是所述目标方向集的子集。
6、 根据权利要求 3所述的方法, 其特征在于, 所述方法还包括: 从所述目标方向集中确定所述基本方向集的方法包括: 从所述目标方向 集中确定基准方向, 选择基准方向起始的 0度到 45度(包含 0度角与 45度 角) 区间作为最小弧度(角度) 区间。 将目标方向集中所有方向经水平、 垂 直翻转与 (或)对角线转置操作变换到该弧度(角度) 区间内, 此时该区间 内的所有方向组成基本方向集。
7、 根据权利要求 3所述的方法, 其特征在于, 所述方法还包括: 从所述目标方向集中确定所述基本方向集的方法包括: 从所述目标方向 集中确定基准方向, 选择基准方向起始的 0度到 90度(包含 0度角与 90度 角) 区间作为最小弧度(角度) 区间。 将目标方向集中所有方向经水平、 垂 直翻转与 (或)对角线转置操作变换到该弧度(角度) 区间内, 此时该区间 内的所有方向组成基本方向集。
8、 根据权利要求 6或 7所述的方法, 其特征在于, 所述方法还包括: 所述基准方向为水平方向或垂直方向。
9、 根据权利要求 1或 2所述的方法, 其特征在于, 所述方法还包括: 所述基本方向与方向性变换矩阵的映射关系表, 记录了基本方向与方向 性变换矩阵的映射关系。
1 0、 根据权利要求 9所述的基本方向与方向性变换矩阵的映射关系表, 其特征在于, 所述方法还包括:
对不可分离变换, 一个基本方向对应着一个不可分离二维方向性变换矩 阵。
1 1、 根据权利要求 9所述的基本方向与方向性变换矩阵的映射关系表, 其特征在于, 所述方法还包括:
对可分离变换, 一个基本方向对应着一个一维的列变换矩阵和一个一维 的行变换矩阵。
12、 一种视频解码的装置, 其特征在于, 包括:
存储单元, 用于存储基本方向和目标方向之间的几何变换和几何逆变换 映射关系表; 该单元还用于存储基本方向与方向性变换矩阵的映射关系表。
方向性逆变换单元, 用于获得当前块的目标方向, 从所述基本方向和目 标方向之间的几何变换和几何逆变换映射关系表中获得对应的基本方向及一 个或一组几何逆变换, 根据所述基本方向从所述基本方向与方向性变换矩阵 的映射关系表中获得方向性变换矩阵, 使用所述方向性变换矩阵, 对解码端 接收到的变换系数矩阵做空间方向性逆变换, 得到块信号。
几何逆变换单元, 用于使用所述一个或一组几何逆变换, 对所述块信号 做几何逆变换, 得到重建块信号。
1 3、 一种视频编码的装置, 其特征在于, 包括:
存储单元, 用于存储基本方向和目标方向之间的几何变换和几何逆变换 映射关系表; 该单元还用于存储基本方向与方向性变换矩阵的映射关系表。
几何变换单元, 用于获得当前块的目标方向, 从所述基本方向和目标方 向之间的几何变换和几何逆变换映射关系表中获得对应的基本方向及一个或 一组几何变换, 使用所述一个或一组几何变换对图像块信号进行几何变换, 得到几何变换后的图像块信号。
方向性变换单元, 用于根据所述基本方向, 从所述基本方向与方向性变 换矩阵的映射关系表中获得对应的方向性变换矩阵, 使用所述方向性变换矩 阵对所述几何变换后的图像块信号做方向性变换, 得到变换系数矩阵。
14、 一种方向性变换的帧间预测视频解码方法, 其特征在于, 包括: 以当前帧间的方向性预测模式为目标方向, 从帧间预测时目标方向与方 向性变换矩阵的映射关系表中获得对应的方向性变换矩阵;
使用所述方向性变换矩阵对解码端接收到的变换系数矩阵做方向性逆变 换, 得到重建块信号。
15、 一种方向性变换的帧间预测视频编码方法, 其特征在于, 包括: 以当前帧间的方向性预测模式为目标方向, 从帧间预测时目标方向与方 向性变换矩阵的映射关系表中获得对应的方向性变换矩阵;
使用所述方向性变换矩阵对图像块信号做方向性变换, 得到变换系数矩 阵。
16、 根据权利要求 14或 15所述的方法, 其特征在于, 所述方法还包括: 所述帧间预测时目标方向与方向性变换矩阵的映射关系表记录了帧间预 测编码和解码下目标方向与方向性变换矩阵的映射关系。
17、 一种帧间预测视频解码的装置, 其特征在于, 包括:
存储单元, 用于存储帧间预测时目标方向与方向性变换矩阵的映射关系 表。
方向性逆变换单元, 用于以当前帧间的方向性预测模式为目标方向, 从 所述帧间预测时目标方向与方向性变换矩阵的映射关系表中获得对应的方向 性变换矩阵, 使用所述方向性变换矩阵对解码端接收到的变换系数矩阵做方 向性逆变换, 得到重建块信号。
18、 一种帧间预测视频编码的装置, 其特征在于, 包括:
存储单元, 用于存储帧间预测时目标方向与方向性变换矩阵的映射关系 表。
方向性变换单元, 用于以当前帧间的方向性预测模式为目标方向, 从所 述帧间预测时目标方向与方向性变换矩阵的映射关系表中获得对应的方向性 变换矩阵, 使用所述方向性变换矩阵对图像块信号做方向性变换, 得到变换 系数矩阵。
PCT/CN2011/072564 2010-04-09 2011-04-09 一种简化方向性变换的方法、装置和系统 WO2011124158A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN 201010147559 CN102215388B (zh) 2010-04-09 2010-04-09 一种简化方向性变换的方法、装置和系统
CN201010147559.9 2010-04-09

Publications (1)

Publication Number Publication Date
WO2011124158A1 true WO2011124158A1 (zh) 2011-10-13

Family

ID=44746485

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2011/072564 WO2011124158A1 (zh) 2010-04-09 2011-04-09 一种简化方向性变换的方法、装置和系统

Country Status (2)

Country Link
CN (1) CN102215388B (zh)
WO (1) WO2011124158A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109587479B (zh) * 2017-09-29 2023-11-10 华为技术有限公司 视频图像的帧间预测方法、装置及编解码器
CN114513661B (zh) * 2022-04-20 2022-09-06 宁波康达凯能医疗科技有限公司 一种基于方向检测的帧内图像模式决策方法与系统

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP3654664B2 (ja) * 1994-08-23 2005-06-02 シャープ株式会社 画像符号化装置及び画像復号装置
WO2008157269A2 (en) * 2007-06-15 2008-12-24 Qualcomm Incorporated Adaptive coding of video block prediction mode
CN101489134A (zh) * 2009-01-16 2009-07-22 华中科技大学 用于视频帧内编码的klt矩阵训练方法

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP3979403B2 (ja) * 2004-06-28 2007-09-19 ソニー株式会社 画像情報処理装置及び画像情報処理方法
WO2008042127A2 (en) * 2006-09-29 2008-04-10 Thomson Licensing Geometric intra prediction
CN101227601B (zh) * 2007-01-15 2011-09-14 飞思卡尔半导体公司 在视频再现中进行几何变换的方法和设备

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP3654664B2 (ja) * 1994-08-23 2005-06-02 シャープ株式会社 画像符号化装置及び画像復号装置
WO2008157269A2 (en) * 2007-06-15 2008-12-24 Qualcomm Incorporated Adaptive coding of video block prediction mode
CN101489134A (zh) * 2009-01-16 2009-07-22 华中科技大学 用于视频帧内编码的klt矩阵训练方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
ZHANG NAN ET AL.: "Spatial Prediction Based Intra-Coding.", 2004 IEEE INTERNATIONAL CONFERENCE ON MULTIMEDIA AND EXPO (ICME)., vol. 1, 2004, pages 97 - 100 *

Also Published As

Publication number Publication date
CN102215388A (zh) 2011-10-12
CN102215388B (zh) 2013-11-06

Similar Documents

Publication Publication Date Title
JP6858277B2 (ja) 方向性イントラ予測コーディング
JP2018511227A (ja) フレーム内符号化方法、フレーム内復号化方法、エンコーダ、及びデコーダ
US11272204B2 (en) Motion compensation method and device, and computer system
US7526134B2 (en) Image processing apparatus, program, recording medium, and data decompression method
TW200843519A (en) Systems and methods for efficient spatial intra predictabilty determination (or assessment)
CN110636313B (zh) 变换、二次变换矩阵训练方法、编码器及相关装置
JP2020529782A (ja) ビデオ圧縮のための変換を構成する方法および装置
WO2020047756A1 (zh) 图像编码方法和装置
JP2012138883A5 (zh)
WO2009097809A1 (zh) 基于自适应块变换的帧内预测方法及装置
WO2020103800A1 (zh) 视频解码方法和视频解码器
BR122021019694B1 (pt) Método de decodificação / codificação de imagem realizado por um aparelho de decodificação / codificação, aparelho de decodificação/codificação para decodificação/codificação de imagem, método e aparelho de transmissão de dados que compreende um fluxo de bits para uma imagem e mídia de armazenamento digital legível por computador não transitória
WO2013064120A1 (zh) 一种变换模式的编解码方法和装置
TW200917851A (en) Enhanced compression in representing non-frame-edge blocks of image frames
WO2012122948A1 (zh) 图像编解码方法、图像编解码设备及网络系统
WO2020038378A1 (zh) 色度块预测方法及装置
WO2021134635A1 (zh) 变换方法、编码器、解码器以及存储介质
WO2011124158A1 (zh) 一种简化方向性变换的方法、装置和系统
TW201228402A (en) Inverse transform calculation method with low complexity
WO2013064109A1 (zh) 一种图像编码、解码的方法和装置
JP4825227B2 (ja) 画像復号装置及び画像処理プログラム
WO2011124163A1 (zh) 视频数据编码、解码方法及装置、变换处理方法及装置
WO2012175033A1 (zh) 图像变换处理方法、设备和系统
WO2019194505A1 (ko) 축소된 회전기반 변환을 이용하여 비디오 신호를 처리하는 방법 및 장치
KR20170058335A (ko) 분리 가능한 변환에 기초하여 적응적으로 비디오 신호를 인코딩 및 디코딩하는 방법 및 장치

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 11765076

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 11765076

Country of ref document: EP

Kind code of ref document: A1