CN104079937B - It is a kind of based on motion vector analysis by H.264 to the quick interframe code-transferring method and transcoding device of HEVC - Google Patents

Abstract

The present invention relates to it is a kind of based on motion vector analysis by H.264 to the quick interframe code-transferring method of HEVC, specific steps include：(1) H.264 code stream is decoded in H.264 decoder terminal, extracts reference frame information, macroblock size, macro block position, the motion vector MV of macro block, the prediction mode of macro block in described H.264 code stream；(2) the motion vector MV of the macro block according to the H.264 code stream obtained in H.264 decoder terminal determines coding unit CU dividing modes；(3) piecemeal according to coding unit CU determines the dividing mode of predicting unit PU；(4) motion vector information by being extracted in H.264 code stream directly predicts the motion vector of predicting unit PU in current HEVC.The quick inter prediction code-transferring method that the present invention is provided makes the relevant information in H.264 code stream maximize the use, and the complexity of HEVC coding sides is reduced well；The time of HEVC codings is reduced, code efficiency is greatly improved.

Description

It is a kind of based on motion vector analysis by H.264 to the quick interframe transcoding side of HEVC Method and transcoding device

Technical field

The present invention relates to it is a kind of based on motion vector analysis by H.264 to the quick interframe code-transferring method and transcoding of HEVC Device, belongs to video code conversion field.

Background technology

Current most widely used video encoding and decoding standard is that H.264/AVC, it is widely used in cable television and satellite The field such as TV, wireless and mobile network, multimedia streaming service.However, popularization and ultra high-definition lattice with HD video The rise of formula, original H.264/AVC standard can not well meet demand of the coding and decoding video in effect and speed.Cause This, International Organization for standardization ITU-T and ISO/IEC propose a kind of high-performance video coding HEVC as Video coding of new generation Standard.More H.264/AVC, HEVC improves 30% to 50% in coding and decoding video aspect of performance, but HEVC is in coding side simultaneously Calculation cost and increasingly complex storage organization higher is paid.Also, nowadays H.264 broader applications in shooting Machine, mobile phone, at a distance of etc. in the video capture device of main flow.HEVC cannot replace status H.264 in a short time, that is to say, that In significant period of time, two kinds of video standards will coexist.Therefore, to that H.26/AVC4 will be helped to the research of HEVC transcodings Fusion and transition between two kinds of video standards, have important practical significance.

H.264/AVC it is one kind conversion to video format to the process of HEVC transcodings, most straightforward approach is by H.264 Decoder cascaded with the encoder of HEVC, H.264 decoded video is encoded in directly inputting HEVC encoders. However, some algorithms due to introducing intensive and complexity high in HEVC encoders, therefore, using in H.264 code stream Entrained partial information is optimized to the algorithm of this part amount of calculation and complexity high, just becomes a class of research Topic.

Difference relative to macro block MB, HEVC in H.264 coding standard is the introduction of adative quadtree structure Code tree unit CTU as prediction, conversion, quantify, the elementary cell of entropy code.One code tree unit CTU is bright including one Degree coding tree block CTB and two chroma coder tree blocks.The quaternary tree syntax of code tree unit CTU formulated brightness belonging to it and The size of chroma coder block CB and position, a usual luminance coding tree block CTB and two chroma coder tree blocks and their correlations Syntax constitute a coding unit CU.The full-size of coding unit CU is the size of the luminance coding tree block CTB belonging to it. The size of coding unit CU or comprising pixel count could be arranged to 64 × 64,32 × 32,16 × 16 and 8 × 8.In general, Coding unit CU sizes are bigger, and compression performance is better.Each coding unit CU includes predicting unit PU associated therewith With converter unit TU.For predicting unit PU, can be divided into 2N × 2N, 2N × N, the symmetry division pattern of N × 2N and 2N × The asymmetric Fractionation regimen of nU, 2N × nD, nL × 2N, nR × 2N.Compared with the H.264 coding unit of middle fixed macroblock size, Divisions of the HEVC to coded image is more flexible.However, while video coding performance is improved, also increasing cataloged procedure Complexity.

H.264, traditional is to cascade the encoder of H.264 decoder and HEVC to the method for HEVC transcodings, directly right H.24 the image that standard decoding is obtained carries out HEVC standard coding.This method has traveled through all processes in coding, complexity It is higher, the information included in H.264 code stream is not made full use of.

Currently for the transcoding research between conventional various video encoding standards more comprehensively, for H.264 to HEVC The research of transcoding is also being carried out.Tamer Shanabieh and Eduardo Peixoto et al. proposes a kind of by linearly sentencing Other function, sets up a mapping between the partitioned mode in characteristic information and HEVC coding standards in H.264 code stream.Utilize The thought of pattern classification, extracts the characteristic information in H.264 code stream, directly determines the coding unit of image in HEVC encoders Dividing mode.This approach reduces the amount of calculation that coding side determines partitioned mode, so as to improve code rate.

Additionally, Dong Zhang and Bin Li et al. propose the spectra calculation method based on rate-distortion optimization model come Predict the macroblock mode of the CU and PU of I picture.In the case where encoding efficiency is reduced less, larger reduces coding side Computation complexity.

Due to H.264 with HEVC encode code stream in, reference frame number, motion vector, minimum movement estimate prediction mode Etc. being identical.Therefore, it can extract reference frame number, macro block position coordinate, macro block (mb) type, motion vector in H.264 code stream Etc. information.And apply in HEVC coding sides, the links such as interframe and infra-frame prediction are optimized, so as to whole transcoding is greatly reduced The complexity of process.

The content of the invention

For deficiency of the prior art, the invention discloses it is a kind of based on motion vector analysis by H.264 to HEVC Quick interframe code-transferring method；

The present invention also disclosed a kind of transcoding device for realizing above-mentioned code-transferring method.

The technical scheme is that：

It is a kind of based on motion vector analysis by H.264 to the quick interframe code-transferring method of HEVC, specific steps include：

(1) H.264 code stream is decoded in H.264 decoder terminal, decoder terminal extract it is described H.264 Reference frame information, macroblock size, macro block position, the motion vector MV of macro block, the prediction mode of macro block in code stream；

(2) the motion vector MV of the macro block according to the H.264 code stream obtained in H.264 decoder terminal determines coding unit CU dividing modes, the i.e. piecemeal of coding unit CU, comprise the following steps that：

A, the initial block size for setting coding unit CU are 64 × 64；Calculate the coding that initial block size is 64 × 64 The number and type of the H.264 substandard macro block included in unit CU；

B, basis extract the motion vector MV of the macro block described in the step of obtaining a in H.264 decoding end, calculate block size By the variance of the motion vector of each macro block included in 64 × 64 coding unit CU is apart from D_mv,

In formula I, D_mvIt is used to represent block size by the motion of each macro block included in 64 × 64 coding unit CU The otherness of vector, whereinWithRepresent respectively and (x is located in H.264 standard_m,y_m) and (x_n,y_n) place m macro blocks and n macro blocks motion vector MV abscissa and ordinate；

C, one threshold value T of setting₁If, D_mvLess than T₁When, then block size need not enter for 64 × 64 coding unit CU Line splitting, block size 64 × 64 now is exactly coding unit CU dividing modes；If D_mvMore than or equal to T₁When, then piecemeal chi Very little is that 64 × 64 coding unit CU enters line splitting treatment, splits into the coding unit CU that 4 block sizes are 32 × 32, is entered Enter step d；

D, the initial block size for setting coding unit CU are 32 × 32, calculate the coding that initial block size is for 32 × 32 The number and type of the H.264 substandard macro block included in unit CU；

E, basis extract the motion vector MV of the macro block described in the step of obtaining d in H.264 decoding end, calculate block size By the variance of the motion vector of each macro block included in 32 × 32 coding unit CU is apart from D_mv,

In formula I, D_mvIt is used to represent block size by the motion of each macro block included in 32 × 32 coding unit CU The otherness of vector, whereinWithRepresent respectively and (x is located in H.264 standard_m,y_m) and (x_n,y_n) place m macro blocks and n macro blocks motion vector MV abscissa and ordinate；

F, one threshold value T of setting₁If, D_mvLess than T₁When, then block size need not enter for 32 × 32 coding unit CU Line splitting, block size 32 × 32 now is exactly coding unit CU dividing modes；If D_mvMore than or equal to T₁When, then piecemeal chi Very little is that 32 × 32 coding unit CU enters line splitting treatment, 4 coding units of block size 16 × 16 is split into, into step g；

G, the initial block size for setting coding unit CU are 16 × 16, calculate the coding that initial block size is for 16 × 16 The number and type of the H.264 substandard macro block included in unit CU；

H, basis extract the motion vector MV of the macro block described in the step of obtaining (g) in H.264 decoding end, calculate piecemeal chi The variance of the motion vector of each macro block included in the very little coding unit CU by 16 × 16 is apart from D_mv,

In formula I, D_mvIt is used to represent block size by the motion of each macro block included in 16 × 16 coding unit CU The otherness of vector, whereinWithRepresent respectively and (x is located in H.264 standard_m,y_m) and (x_n,y_n) place m macro blocks and n macro blocks motion vector MV abscissa and ordinate；

I, one threshold value T of setting₁If, D_mvLess than T₁When, then block size need not enter for 16 × 16 coding unit CU Line splitting, block size 16 × 16 now is exactly coding unit CU dividing modes；If D_mvMore than or equal to T₁When, then piecemeal chi Very little is that 16 × 16 coding unit CU enters line splitting treatment, splits into 4 coding units of block size 8 × 8, piecemeal now Size 8 × 8 is exactly coding unit CU dividing modes；

Work as D_mvLess than T₁When, i.e., block size is sweared by the motion of each macro block included in 64 × 64 coding unit CU Amount difference is little, without entering line splitting；D_mvMore than or equal to T₁When, i.e., block size in 64 × 64 coding unit CU by including Each macro block motion vector difference it is larger, then enter line splitting treatment；

(3) piecemeal according to coding unit CU determines the dividing mode of predicting unit PU, according to coding unit CU division sides Formula determines the dividing mode of predicting unit PU, and each block size is that the coding unit CU of 2N × 2N has 7 kinds of predicting unit PU Dividing mode, wherein, described N=4,8,16,32；The dividing mode of 7 kinds of described predicting unit PU is respectively 2N × 2N, 2N × N, the symmetry division pattern of N × 2N and 2N × nU, 2N × nD, nL × 2N, the asymmetric Fractionation regimen of nR × 2N, wherein, 7 kinds of described dividing modes of predicting unit PU are international standard, and U is up, and D is down, and R is right, and L is left, and n is non- The mark of symmetric pattern, i.e. 2N × nU with n, 2N × nD, nL × 2N, nR × 2N are asymmetric mode, without the 2N with n × 2N, 2N × N, N × 2N are symmetric pattern；The asymmetric Fractionation regimens of described 2N × nU are from top to bottom 1:3 ratio cut partition is pre- Survey unit PU；The asymmetric Fractionation regimens of described 2N × nD are from top to bottom 3:1 ratio cut partition predicting unit PU；Described nL The asymmetric Fractionation regimens of × 2N are from left to right 1:3 ratio cut partition predicting unit PU；The asymmetric segmentation moulds of described nR × 2N Formula is from left to right 3:1 ratio cut partition predicting unit PU；Comprise the following steps that：

If the block size of j, coding unit CU is 8 × 8, the coding unit CU is divided into 4 block sizes It is 4 × 4 fritter；Otherwise, described coding unit CU is divided into 16 block size identical fritters；

K, according to the motion vector information obtained in H.264 code stream, extract the motion of each fritter in coding unit CU Vector, and calculate the phase of the motion vector of fritter described in each；

By the side of the phase of each fritter in coding unit CU under following formula II 7 kinds of dividing modes of predicting unit PU of calculating Difference sum D_ψ, described 7 kinds of dividing modes of predicting unit PU are 2N × 2N, 2N × N, the symmetry division pattern of N × 2N and 2N × nU, 2N × nD, nL × 2N, the asymmetric Fractionation regimen of nR × 2N；

D_ψ=D_ψ1+D_ψ2 ⅱ

In formula II, ψ₁,ψ₂Two segmented areas in predicting unit PU dividing modes are represented respectively； Wherein, (i, j) is ψ₁The transverse and longitudinal coordinate of piecemeal；(i ', j ') is ψ₂The transverse and longitudinal coordinate of piecemeal；

Compare D in the case of 7 kinds of described dividing modes of predicting unit PU_ψ, choose and cause D_ψPredicting unit when value is minimum The dividing mode of PU is defined as the dividing mode of predicting unit PU；

(4) motion vector information by being extracted in H.264 code stream directly predicts predicting unit PU in current HEVC Motion vector, concretely comprise the following steps：

L, the motion vector according to the H.264 code stream extracted in H.264 decoder terminal in step (l), look for respectively The best matching blocks in reference to frame unit corresponding to each fritter included in the predicting unit PU for determining dividing mode； Described best matching blocks refer to the reference frame unit closest with each fritter included in described predicting unit PU；It is logical Cross following formula III and calculate the variance with reference to each fritter included in frame unit and predicting unit PU respectively apart from D,

In formula III,It is t-th position of fritter in predicting unit PU；It is t-th fritter correspondence in PU units Best matching blocks position；

Wherein, with the variance of each fritter that includes in described predicting unit PU apart from the minimum reference frame unit of sum As described best matching blocks；

If m, obtaining a best matching blocks by step (l), predicting unit PU is directly calculated optimal with described Skew between match block is the motion vector of described predicting unit PU；If obtaining two or more by step (l) Best matching blocks, then calculate and determine the predicting unit PU of dividing mode and asking for described two or more best matching blocks respectively With absolute error criterion SAD, the minimum best matching blocks of sad value are chosen as best matching blocks, directly calculate predicting unit PU Skew between described best matching blocks is the motion vector of described predicting unit PU.

It is a kind of realize based on motion vector analysis by H.264 to HEVC quick interframe code-transferring method transcoding device, Including：H.264 decoder module, motion vector information extraction module, storage module, calculate processing module and HEVC coding modules, The described motion vector information extraction module of described H.264 decoder module connection, described motion vector information extraction module Connect the storage module, the described calculating processing module of described storage module connection, described calculating processing module connection Described HEVC coding modules.

H.264 code stream is carried out in H.264 decoder terminal according to currently preferred, described H.264 decoder module Decoding；Described motion vector information extraction module is used to be wrapped in H.264 decoder terminal extracts described H.264 code stream Reference frame information, macroblock size, macro block position, the motion vector MV of macro block, the prediction mode of macro block for containing；Described storage mould Block, for store motion vector information extraction module extract described in H.264 code stream in the reference frame information, the macro block that include Size, macro block position, the motion vector MV of macro block, the prediction mode of macro block；Described calculating processing module is used for calculation code The variance distance of the motion vector of each macro block included in unit CU, for the phase of each fritter in calculation code unit CU The variance sum of position, for calculating the variance of each fritter included in reference frame cell distance and predicting unit PU apart from it Be additionally operable to calculate the skew between predicting unit PU and described best matching blocks；Described HEVC coding modules, for root According to the data result that the motion vector information and calculating processing module that are extracted from H.264 code stream are obtained, present frame is carried out Optimized Coding Based.

Beneficial effects of the present invention are：

1st, the quick inter prediction code-transferring method that the present invention is provided makes the relevant information in H.264 code stream obtain maximized Utilize, the complexity of HEVC coding sides is reduced well.

2nd, the present invention can reduce the time of HEVC codings, greatly improve code efficiency.

Brief description of the drawings：

Fig. 1 be the present invention it is a kind of realizes be based on motion vector analysis by H.264 to the quick interframe code-transferring method of HEVC Transcoding device structure principle chart.

Specific embodiment

With reference to Figure of description and embodiment, the invention will be further described, and following specific embodiments are only used for Illustrate the present invention rather than limitation the scope of the present invention.

Embodiment 1

It is a kind of based on motion vector analysis by H.264 to the quick interframe code-transferring method of HEVC, specific steps include：

(1) H.264 code stream is decoded in H.264 decoder terminal, decoder terminal extract it is described H.264 Reference frame information, macroblock size, macro block position, the motion vector MV of macro block, the prediction mode of macro block in code stream；

(2) the motion vector MV of the macro block according to the H.264 code stream obtained in H.264 decoder terminal determines coding unit CU dividing modes, the i.e. piecemeal of coding unit CU, comprise the following steps that：

A, the initial block size for setting coding unit CU are 64 × 64；Calculate the coding that initial block size is 64 × 64 The number and type of the H.264 substandard macro block included in unit CU；

B, basis extract the motion vector MV of the macro block described in the step of obtaining a in H.264 decoding end, calculate block size By the variance of the motion vector of each macro block included in 64 × 64 coding unit CU is apart from D_mv,

In formula I, D_mvIt is used to represent block size by the motion of each macro block included in 64 × 64 coding unit CU The otherness of vector, whereinWithRepresent respectively and (x is located in H.264 standard_m,y_m) and (x_n,y_n) place m macro blocks and n macro blocks motion vector MV abscissa and ordinate；

C, one threshold value T of setting₁If, D_mvLess than T₁When, then block size need not enter for 64 × 64 coding unit CU Line splitting, block size 64 × 64 now is exactly coding unit CU dividing modes；If D_mvMore than or equal to T₁When, then piecemeal chi Very little is that 64 × 64 coding unit CU enters line splitting treatment, splits into the coding unit CU that 4 block sizes are 32 × 32, is entered Enter step d；

D, the initial block size for setting coding unit CU are 32 × 32, calculate the coding that initial block size is for 32 × 32 The number and type of the H.264 substandard macro block included in unit CU；

E, basis extract the motion vector MV of the macro block described in the step of obtaining d in H.264 decoding end, calculate block size By the variance of the motion vector of each macro block included in 32 × 32 coding unit CU is apart from D_mv,

In formula I, D_mvIt is used to represent block size by the motion of each macro block included in 32 × 32 coding unit CU The otherness of vector, whereinWithRepresent respectively and (x is located in H.264 standard_m,y_m) and (x_n,y_n) place m macro blocks and n macro blocks motion vector MV abscissa and ordinate；

F, one threshold value T of setting₁If, D_mvLess than T₁When, then block size need not enter for 32 × 32 coding unit CU Line splitting, block size 32 × 32 now is exactly coding unit CU dividing modes；If D_mvMore than or equal to T₁When, then piecemeal chi Very little is that 32 × 32 coding unit CU enters line splitting treatment, 4 coding units of block size 16 × 16 is split into, into step g；

G, the initial block size for setting coding unit CU are 16 × 16, calculate the coding that initial block size is for 16 × 16 The number and type of the H.264 substandard macro block included in unit CU；

H, basis extract the motion vector MV of the macro block described in the step of obtaining (g) in H.264 decoding end, calculate piecemeal chi The variance of the motion vector of each macro block included in the very little coding unit CU by 16 × 16 is apart from D_mv,

In formula I, D_mvIt is used to represent block size by the motion of each macro block included in 16 × 16 coding unit CU The otherness of vector, whereinWithRepresent respectively and (x is located in H.264 standard_m,y_m) and (x_n,y_n) place m macro blocks and n macro blocks motion vector MV abscissa and ordinate；

I, one threshold value T of setting₁If, D_mvLess than T₁When, then block size need not enter for 16 × 16 coding unit CU Line splitting, block size 16 × 16 now is exactly coding unit CU dividing modes；If D_mvMore than or equal to T₁When, then piecemeal chi Very little is that 16 × 16 coding unit CU enters line splitting treatment, splits into 4 coding units of block size 8 × 8, piecemeal now Size 8 × 8 is exactly coding unit CU dividing modes；

Work as D_mvLess than T₁When, i.e., block size is sweared by the motion of each macro block included in 64 × 64 coding unit CU Amount difference is little, without entering line splitting；D_mvMore than or equal to T₁When, i.e., block size in 64 × 64 coding unit CU by including Each macro block motion vector difference it is larger, then enter line splitting treatment；

(3) piecemeal according to coding unit CU determines the dividing mode of predicting unit PU, according to coding unit CU division sides Formula determines the dividing mode of predicting unit PU, and each block size is that the coding unit CU of 2N × 2N has 7 kinds of predicting unit PU Dividing mode, wherein, described N=4,8,16,32；The dividing mode of 7 kinds of described predicting unit PU is respectively 2N × 2N, 2N × N, the symmetry division pattern of N × 2N and 2N × nU, 2N × nD, nL × 2N, the asymmetric Fractionation regimen of nR × 2N, wherein, 7 kinds of described dividing modes of predicting unit PU are international standard, and U is up, and D is down, and R is right, and L is left, and n is non- The mark of symmetric pattern, i.e. 2N × nU with n, 2N × nD, nL × 2N, nR × 2N are asymmetric mode, without the 2N with n × 2N, 2N × N, N × 2N are symmetric pattern；The asymmetric Fractionation regimens of described 2N × nU are from top to bottom 1:3 ratio cut partition is pre- Survey unit PU；The asymmetric Fractionation regimens of described 2N × nD are from top to bottom 3:1 ratio cut partition predicting unit PU；Described nL The asymmetric Fractionation regimens of × 2N are from left to right 1:3 ratio cut partition predicting unit PU；The asymmetric segmentation moulds of described nR × 2N Formula is from left to right 3:1 ratio cut partition predicting unit PU；Comprise the following steps that：

If the block size of j, coding unit CU is 8 × 8, the coding unit CU is divided into 4 block sizes It is 4 × 4 fritter；Otherwise, described coding unit CU is divided into 16 block size identical fritters；

K, according to the motion vector information obtained in H.264 code stream, extract the motion of each fritter in coding unit CU Vector, and calculate the phase of the motion vector of fritter described in each；

By the side of the phase of each fritter in coding unit CU under following formula II 7 kinds of dividing modes of predicting unit PU of calculating Difference sum D_ψ, described 7 kinds of dividing modes of predicting unit PU are 2N × 2N, 2N × N, the symmetry division pattern of N × 2N and 2N × nU, 2N × nD, nL × 2N, the asymmetric Fractionation regimen of nR × 2N；

D_ψ=D_ψ1+D_ψ2 ⅱ

In formula II, ψ₁,ψ₂Two segmented areas in predicting unit PU dividing modes are represented respectively； Wherein, (i, j) is ψ₁The transverse and longitudinal coordinate of piecemeal；(i ', j ') is ψ₂The transverse and longitudinal coordinate of piecemeal；

Compare D in the case of 7 kinds of described dividing modes of predicting unit PU_ψ, choose and cause D_ψPredicting unit when value is minimum The dividing mode of PU is defined as the dividing mode of predicting unit PU；

(4) motion vector information by being extracted in H.264 code stream directly predicts predicting unit PU in current HEVC Motion vector, concretely comprise the following steps：

L, the motion vector according to the H.264 code stream extracted in H.264 decoder terminal in step (l), look for respectively The best matching blocks in reference to frame unit corresponding to each fritter included in the predicting unit PU for determining dividing mode； Described best matching blocks refer to the reference frame unit closest with each fritter included in described predicting unit PU；It is logical Cross following formula III and calculate the variance with reference to each fritter included in frame unit and predicting unit PU respectively apart from D,

In formula III,It is t-th position of fritter in predicting unit PU；It is t-th fritter correspondence in PU units Best matching blocks position；

Wherein, with the variance of each fritter that includes in described predicting unit PU apart from the minimum reference frame unit of sum As described best matching blocks；

M, a best matching blocks are obtained by step (l), then directly calculate predicting unit PU and described best match Skew between block is the motion vector of described predicting unit PU.

Embodiment 2

It is a kind of based on motion vector analysis by H.264 to the quick interframe code-transferring method of HEVC, specific steps include：

(1) H.264 code stream is decoded in H.264 decoder terminal, decoder terminal extract it is described H.264 Reference frame information, macroblock size, macro block position, the motion vector MV of macro block, the prediction mode of macro block in code stream；

(2) the motion vector MV of the macro block according to the H.264 code stream obtained in H.264 decoder terminal determines coding unit CU dividing modes, the i.e. piecemeal of coding unit CU, comprise the following steps that：

A, the initial block size for setting coding unit CU are 64 × 64；Calculate the coding that initial block size is 64 × 64 The number and type of the H.264 substandard macro block included in unit CU；

B, basis extract the motion vector MV of the macro block described in the step of obtaining a in H.264 decoding end, calculate block size By the variance of the motion vector of each macro block included in 64 × 64 coding unit CU is apart from D_mv,

In formula I, D_mvIt is used to represent block size by the motion of each macro block included in 64 × 64 coding unit CU The otherness of vector, whereinWithRepresent respectively and (x is located in H.264 standard_m,y_m) and (x_n,y_n) place m macro blocks and n macro blocks motion vector MV abscissa and ordinate；

C, one threshold value T of setting₁If, D_mvLess than T₁When, then block size need not enter for 64 × 64 coding unit CU Line splitting, block size 64 × 64 now is exactly coding unit CU dividing modes；If D_mvMore than or equal to T₁When, then piecemeal chi Very little is that 64 × 64 coding unit CU enters line splitting treatment, splits into the coding unit CU that 4 block sizes are 32 × 32, is entered Enter step d；

D, the initial block size for setting coding unit CU are 32 × 32, calculate the coding that initial block size is for 32 × 32 The number and type of the H.264 substandard macro block included in unit CU；

E, basis extract the motion vector MV of the macro block described in the step of obtaining d in H.264 decoding end, calculate block size By the variance of the motion vector of each macro block included in 32 × 32 coding unit CU is apart from D_mv,

In formula I, D_mvIt is used to represent block size by the motion of each macro block included in 32 × 32 coding unit CU The otherness of vector, whereinWithRepresent respectively and (x is located in H.264 standard_m,y_m) and (x_n,y_n) place m macro blocks and n macro blocks motion vector MV abscissa and ordinate；

F, one threshold value T of setting₁If, D_mvLess than T₁When, then block size need not enter for 32 × 32 coding unit CU Line splitting, block size 32 × 32 now is exactly coding unit CU dividing modes；If D_mvMore than or equal to T₁When, then piecemeal chi Very little is that 32 × 32 coding unit CU enters line splitting treatment, 4 coding units of block size 16 × 16 is split into, into step g；

G, the initial block size for setting coding unit CU are 16 × 16, calculate the coding that initial block size is for 16 × 16 The number and type of the H.264 substandard macro block included in unit CU；

H, basis extract the motion vector MV of the macro block described in the step of obtaining (g) in H.264 decoding end, calculate piecemeal chi The variance of the motion vector of each macro block included in the very little coding unit CU by 16 × 16 is apart from D_mv,

In formula I, D_mvIt is used to represent block size by the motion of each macro block included in 16 × 16 coding unit CU The otherness of vector, whereinWithRepresent respectively and (x is located in H.264 standard_m,y_m) and (x_n,y_n) place m macro blocks and n macro blocks motion vector MV abscissa and ordinate；

I, one threshold value T of setting₁If, D_mvLess than T₁When, then block size need not enter for 16 × 16 coding unit CU Line splitting, block size 16 × 16 now is exactly coding unit CU dividing modes；If D_mvMore than or equal to T₁When, then piecemeal chi Very little is that 16 × 16 coding unit CU enters line splitting treatment, splits into 4 coding units of block size 8 × 8, piecemeal now Size 8 × 8 is exactly coding unit CU dividing modes；

Work as D_mvLess than T₁When, i.e., block size is sweared by the motion of each macro block included in 64 × 64 coding unit CU Amount difference is little, without entering line splitting；D_mvMore than or equal to T₁When, i.e., block size in 64 × 64 coding unit CU by including Each macro block motion vector difference it is larger, then enter line splitting treatment；

(3) piecemeal according to coding unit CU determines the dividing mode of predicting unit PU, according to coding unit CU division sides Formula determines the dividing mode of predicting unit PU, and each block size is that the coding unit CU of 2N × 2N has 7 kinds of predicting unit PU Dividing mode, wherein, described N=4,8,16,32；The dividing mode of 7 kinds of described predicting unit PU is respectively 2N × 2N, 2N × N, the symmetry division pattern of N × 2N and 2N × nU, 2N × nD, nL × 2N, the asymmetric Fractionation regimen of nR × 2N, wherein, 7 kinds of described dividing modes of predicting unit PU are international standard, and U is up, and D is down, and R is right, and L is left, and n is non- The mark of symmetric pattern, i.e. 2N × nU with n, 2N × nD, nL × 2N, nR × 2N are asymmetric mode, without the 2N with n × 2N, 2N × N, N × 2N are symmetric pattern；The asymmetric Fractionation regimens of described 2N × nU are from top to bottom 1:3 ratio cut partition is pre- Survey unit PU；The asymmetric Fractionation regimens of described 2N × nD are from top to bottom 3:1 ratio cut partition predicting unit PU；Described nL The asymmetric Fractionation regimens of × 2N are from left to right 1:3 ratio cut partition predicting unit PU；The asymmetric segmentation moulds of described nR × 2N Formula is from left to right 3:1 ratio cut partition predicting unit PU；Comprise the following steps that：

If the block size of j, coding unit CU is 8 × 8, the coding unit CU is divided into 4 block sizes It is 4 × 4 fritter；Otherwise, described coding unit CU is divided into 16 block size identical fritters；

K, according to the motion vector information obtained in H.264 code stream, extract the motion of each fritter in coding unit CU Vector, and calculate the phase of the motion vector of fritter described in each；

By the side of the phase of each fritter in coding unit CU under following formula II 7 kinds of dividing modes of predicting unit PU of calculating Difference sum D_ψ, described 7 kinds of dividing modes of predicting unit PU are 2N × 2N, 2N × N, the symmetry division pattern of N × 2N and 2N × nU, 2N × nD, nL × 2N, the asymmetric Fractionation regimen of nR × 2N；

D_ψ=D_ψ1+D_ψ2 ⅱ

In formula II, ψ₁,ψ₂Two segmented areas in predicting unit PU dividing modes are represented respectively； Wherein, (i, j) is ψ₁The transverse and longitudinal coordinate of piecemeal；(i ', j ') is ψ₂The transverse and longitudinal coordinate of piecemeal；

Compare D in the case of 7 kinds of described dividing modes of predicting unit PU_ψ, choose and cause D_ψPredicting unit when value is minimum The dividing mode of PU is defined as the dividing mode of predicting unit PU；

(4) motion vector information by being extracted in H.264 code stream directly predicts predicting unit PU in current HEVC Motion vector, concretely comprise the following steps：

L, the motion vector according to the H.264 code stream extracted in H.264 decoder terminal in step (l), look for respectively The best matching blocks in reference to frame unit corresponding to each fritter included in the predicting unit PU for determining dividing mode； Described best matching blocks refer to the reference frame unit closest with each fritter included in described predicting unit PU；It is logical Cross following formula III and calculate the variance with reference to each fritter included in frame unit and predicting unit PU respectively apart from D,

In formula III,It is t-th position of fritter in predicting unit PU；It is t-th fritter correspondence in PU units Best matching blocks position；

Wherein, with the variance of each fritter that includes in described predicting unit PU apart from the minimum reference frame unit of sum As described best matching blocks；

M, two or more best matching blocks are obtained by step (l), then calculate the prediction list for determining dividing mode respectively The first PU and summation absolute error criterion SAD of described two or more best matching blocks, chooses the minimum best match of sad value Block is described prediction as best matching blocks, the skew directly calculated between predicting unit PU and described best matching blocks The motion vector of unit PU.

Embodiment 3

It is a kind of realize based on motion vector analysis by H.264 to HEVC quick interframe code-transferring method transcoding device, Including：H.264 decoder module, motion vector information extraction module, storage module, calculate processing module and HEVC coding modules, The described motion vector information extraction module of described H.264 decoder module connection, described motion vector information extraction module Connect the storage module, the described calculating processing module of described storage module connection, described calculating processing module connection Described HEVC coding modules.

Described H.264 decoder module is decoded in H.264 decoder terminal to H.264 code stream；Described motion arrow Amount information extraction modules are used to extract the reference frame information, grand that is included in described H.264 code stream in H.264 decoder terminal Block size, macro block position, the motion vector MV of macro block, the prediction mode of macro block；Described storage module, for storing motion arrow It is the reference frame information that is included in the amount described H.264 code stream that extracts of information extraction modules, macroblock size, macro block position, grand The motion vector MV of block, the prediction mode of macro block；Described calculating processing module is used for what is included in calculation code unit CU The variance distance of the motion vector of each macro block, for the variance sum of the phase of each fritter in calculation code unit CU, uses The variance of each fritter included in reference frame cell distance and predicting unit PU is calculated is additionally operable to calculate prediction apart from sum Skew between unit PU and described best matching blocks；Described HEVC coding modules, carry for basis from H.264 code stream The data result that the motion vector information and calculating processing module for taking are obtained, coding is optimized to present frame.

Claims (3)

1. it is a kind of based on motion vector analysis by H.264 to the quick interframe code-transferring method of HEVC, it is characterised in that specific step Suddenly include：

(1) H.264 code stream is decoded in H.264 decoder terminal, described H.264 code stream is extracted in decoder terminal Middle reference frame information, macroblock size, macro block position, the motion vector MV of macro block, the prediction mode of macro block；

(2) the motion vector MV of the macro block according to the H.264 code stream obtained in H.264 decoder terminal determines that coding unit CU draws Divide mode, the i.e. piecemeal of coding unit CU, comprise the following steps that：

A, the initial block size for setting coding unit CU are 64 × 64；Calculate the coding unit that initial block size is 64 × 64 The number and type of the H.264 substandard macro block included in CU；

B, basis extract the motion vector MV of the macro block described in the step of obtaining a in H.264 decoding end, and it is 64 to calculate block size The variance of the motion vector of each macro block included in × 64 coding unit CU is apart from D_mv,

$D_{mv}=\sqrt{{({\mathrm{MV}}_{x_m}-{\mathrm{MV}}_{x_n})}^2+{({\mathrm{MV}}_{y_m}-{\mathrm{MV}}_{y_n})}^2}---i$

In formula I, D_mvIt is used to represent block size by the motion vector of each macro block included in 64 × 64 coding unit CU Otherness, whereinWithRepresent respectively and (x is located in H.264 standard_m,y_m) and (x_n,y_n) The abscissa and ordinate of the m macro blocks at place and the motion vector MV of n macro blocks；

C, one threshold value T of setting₁If, D_mvLess than T₁When, then block size need not be divided for 64 × 64 coding unit CU Split, block size 64 × 64 now is exactly coding unit CU dividing modes；If D_mvMore than or equal to T₁When, then block size is 64 × 64 coding unit CU enters line splitting treatment, the coding unit CU that 4 block sizes are 32 × 32 is split into, into step Rapid d；

D, the initial block size for setting coding unit CU are 32 × 32, calculate the coding unit that initial block size is for 32 × 32 The number and type of the H.264 substandard macro block included in CU；

E, basis extract the motion vector MV of the macro block described in the step of obtaining d in H.264 decoding end, and it is 32 to calculate block size The variance of the motion vector of each macro block included in × 32 coding unit CU is apart from D_mv,

$D_{mv}=\sqrt{{({\mathrm{MV}}_{x_m}-{\mathrm{MV}}_{x_n})}^2+{({\mathrm{MV}}_{y_m}-{\mathrm{MV}}_{y_n})}^2}---i$

In formula I, D_mvIt is used to represent block size by the motion vector of each macro block included in 32 × 32 coding unit CU Otherness, whereinWithRepresent respectively and (x is located in H.264 standard_m,y_m) and (x_n,y_n) The abscissa and ordinate of the m macro blocks at place and the motion vector MV of n macro blocks；

F, one threshold value T of setting₁If, D_mvLess than T₁When, then block size need not be divided for 32 × 32 coding unit CU Split, block size 32 × 32 now is exactly coding unit CU dividing modes；If D_mvMore than or equal to T₁When, then block size is 32 × 32 coding unit CU enters line splitting treatment, 4 coding units of block size 16 × 16 is split into, into step g；

G, the initial block size for setting coding unit CU are 16 × 16, calculate the coding unit that initial block size is for 16 × 16 The number and type of the H.264 substandard macro block included in CU；

H, basis extract the motion vector MV of the macro block described in the step of obtaining g in H.264 decoding end, and it is 16 to calculate block size The variance of the motion vector of each macro block included in × 16 coding unit CU is apart from D_mv,

$D_{mv}=\sqrt{{({\mathrm{MV}}_{x_m}-{\mathrm{MV}}_{x_n})}^2+{({\mathrm{MV}}_{y_m}-{\mathrm{MV}}_{y_n})}^2}---i$

In formula I, D_mvIt is used to represent block size by the motion vector of each macro block included in 16 × 16 coding unit CU Otherness, whereinWithRepresent respectively and (x is located in H.264 standard_m,y_m) and (x_n,y_n) The abscissa and ordinate of the m macro blocks at place and the motion vector MV of n macro blocks；

I, one threshold value T of setting₁If, D_mvLess than T₁When, then block size need not be divided for 16 × 16 coding unit CU Split, block size 16 × 16 now is exactly coding unit CU dividing modes；If D_mvMore than or equal to T₁When, then block size is 16 × 16 coding unit CU enters line splitting treatment, splits into 4 coding units of block size 8 × 8, block size now 8 × 8 is exactly coding unit CU dividing modes；

(3) piecemeal according to coding unit CU determines the dividing mode of predicting unit PU, true according to coding unit CU dividing modes Determine the dividing mode of predicting unit PU, each block size there are 7 kinds of divisions of predicting unit PU for the coding unit CU of 2N × 2N Mode, wherein, described N=4,8,16,32；The dividing mode of 7 kinds of described predicting unit PU be respectively 2N × 2N, 2N × N, The symmetry division pattern and 2N × nU of N × 2N, 2N × nD, nL × 2N, the asymmetric Fractionation regimen of nR × 2N, specific steps are such as Under：

If the block size of j, coding unit CU is 8 × 8, by the coding unit CU be divided into 4 block sizes for 4 × 4 fritter；Otherwise, described coding unit CU is divided into 16 block size identical fritters；

K, according to the motion vector information obtained in H.264 code stream, extract the motion vector of each fritter in coding unit CU, And calculate the phase of the motion vector of fritter described in each

By following formula II calculate the phase of each fritter in coding unit CU under 7 kinds of dividing modes of predicting unit PU variance it And D_ψ, described 7 kinds of dividing modes of predicting unit PU for 2N × 2N, 2N × N, the symmetry division pattern of N × 2N and 2N × The asymmetric Fractionation regimen of nU, 2N × nD, nL × 2N, nR × 2N；

$D_{\mathrm{\ψ}}=D_{{\mathrm{\ψ}}_1}+D_{{\mathrm{\ψ}}_2}---ii$

In formula II, ψ₁,ψ₂Two segmented areas in predicting unit PU dividing modes are represented respectively； Wherein, (i, j) is ψ₁The transverse and longitudinal coordinate of piecemeal；(i ', j ') is ψ₂The transverse and longitudinal coordinate of piecemeal；

Compare D in the case of 7 kinds of described dividing modes of predicting unit PU_ψ, choose and cause D_ψPredicting unit PU when value is minimum Dividing mode is defined as the dividing mode of predicting unit PU；

(4) motion vector information by being extracted in H.264 code stream directly predicts the fortune of predicting unit PU in current HEVC Dynamic vector, concretely comprises the following steps：

L, the motion vector according to the H.264 code stream extracted in H.264 decoder terminal in step (1), find really respectively The best matching blocks in reference to frame unit corresponding to each fritter included in the predicting unit PU for determining dividing mode；It is described Best matching blocks refer to the reference frame unit closest with each fritter included in described predicting unit PU；Under Formula III calculates the variance with reference to each fritter included in frame unit and predicting unit PU apart from D respectively：

In formula III,It is t-th position of fritter in predicting unit PU；For t-th fritter is corresponding optimal in PU units The position of match block；

Wherein, it is apart from the minimum reference frame unit of sum with the variance of each fritter that includes in described predicting unit PU Described best matching blocks；

If m, obtaining a best matching blocks by step l, predicting unit PU and described best matching blocks are directly calculated Between skew be the motion vector of described predicting unit PU；If obtaining two or more best match by step l Block, then calculate and determine that the predicting unit PU of dividing mode is definitely missed with the summation of described two or more best matching blocks respectively Difference criterion SAD, chooses the minimum best matching blocks of sad value as best matching blocks, directly calculating predicting unit PU with it is described Skew between best matching blocks is the motion vector of described predicting unit PU.

2. it is a kind of realize described in claim 1 based on motion vector analysis by H.264 to the quick interframe transcoding side of HEVC The transcoding device of method, it is characterised in that a kind of to realize based on motion vector analysis by H.264 to the quick interframe transcoding of HEVC The transcoding device of method, including：H.264 decoder module, motion vector information extraction module, storage module, calculate processing module With HEVC coding modules, the described motion vector information extraction module of described H.264 decoder module connection, described motion arrow Amount information extraction modules connect the storage module, the described calculating processing module of described storage module connection, described meter Calculate the described HEVC coding modules of processing module connection.

3. it is according to claim 2 it is a kind of realize described in claim 1 based on motion vector analysis by H.264 to The transcoding device of the quick interframe code-transferring method of HEVC, it is characterised in that described H.264 decoder module is in H.264 decoder H.264 code stream is decoded terminal-pair；Described motion vector information extraction module is used to be extracted in H.264 decoder terminal The reference frame information that is included in described H.264 code stream, macroblock size, macro block position, the motion vector MV of macro block, macro block it is pre- Survey mode；Described storage module, for store motion vector information extraction module extract described in H.264 code stream in wrap Reference frame information, macroblock size, macro block position, the motion vector MV of macro block, the prediction mode of macro block for containing；At described calculating Reason module is used for the variance distance of the motion vector of each macro block included in calculation code unit CU, for calculation code list The variance sum of the phase of each fritter in first CU, for calculating each that include in reference frame cell distance and predicting unit PU The variance of fritter is additionally operable to calculate the skew between predicting unit PU and described best matching blocks apart from sum；Described The number that HEVC coding modules, the motion vector information extracted from H.264 code stream for basis and calculating processing module are obtained According to result, coding is optimized to present frame.