CN106507106B - Video interprediction encoding method based on reference plate - Google Patents

Video interprediction encoding method based on reference plate Download PDF

Info

Publication number
CN106507106B
CN106507106B CN201610979281.9A CN201610979281A CN106507106B CN 106507106 B CN106507106 B CN 106507106B CN 201610979281 A CN201610979281 A CN 201610979281A CN 106507106 B CN106507106 B CN 106507106B
Authority
CN
China
Prior art keywords
reference plate
ctu
mrow
current
msub
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201610979281.9A
Other languages
Chinese (zh)
Other versions
CN106507106A (en
Inventor
吴枫
马常月
刘�东
彭秀莲
王婷婷
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Science and Technology of China USTC
Original Assignee
University of Science and Technology of China USTC
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 University of Science and Technology of China USTC filed Critical University of Science and Technology of China USTC
Priority to CN201610979281.9A priority Critical patent/CN106507106B/en
Publication of CN106507106A publication Critical patent/CN106507106A/en
Application granted granted Critical
Publication of CN106507106B publication Critical patent/CN106507106B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102—Methods 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
    • H04N19/103—Selection of coding mode or of prediction mode
    • H04N19/105—Selection of the reference unit for prediction within a chosen coding or prediction mode, e.g. adaptive choice of position and number of pixels used for prediction
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
    • H04N19/136—Incoming video signal characteristics or properties
    • H04N19/137—Motion inside a coding unit, e.g. average field, frame or block difference
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
    • H04N19/146—Data rate or code amount at the encoder output
    • H04N19/149—Data rate or code amount at the encoder output by estimating the code amount by means of a model, e.g. mathematical model or statistical model
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
    • H04N19/154—Measured or subjectively estimated visual quality after decoding, e.g. measurement of distortion
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
    • H04N19/157—Assigned coding mode, i.e. the coding mode being predefined or preselected to be further used for selection of another element or parameter
    • H04N19/159—Prediction type, e.g. intra-frame, inter-frame or bidirectional frame prediction

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Physics & Mathematics (AREA)
  • Algebra (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)

Abstract

The invention discloses a kind of video interprediction encoding method based on reference plate, reference content management and prioritization scheme of this method based on reference plate, can collect more information under conditions of DPB capacity is limited so that present encoding block can more maximum probability find match block.Finally, overall effect is that the efficiency of Video coding is effectively improved in the case where being slightly increased coding and decoding end complexity.

Description

Video interprediction encoding method based on reference plate
Technical field
The present invention relates to technical field of video coding, more particularly to a kind of video interprediction encoding side based on reference plate Method.
Background technology
In recent years, with the fast development of internet, the application demand for video in internet is more and more, and regards The data volume of frequency is very big, in order to be transmitted in the limited internet of bandwidth it may first have to solve the problems, such as to be exactly video compress Encoded question.
The video encoding standard formulated belongs to mixed video coding framework.So-called hybrid video coding, typically by Following sections form：Predict (Prediction), conversion (Transform), quantization (Quantization) and entropy code (Entropy Coding).Wherein prediction is generally divided into infra-frame prediction and inter prediction.Regarding for intra prediction mode can only be selected Frequency frame is referred to as I frames, can use intra prediction mode or be referred to as P frames or B frames using the frame of video of inter-frame forecast mode. Inter-frame forecast mode by the use of other reconstructed frames pixel as reference, present frame is predicted, so as to remove the time Redundancy.Compared to infra-frame prediction, the accuracy rate of general inter prediction is higher, and role is bigger during video compress. The efficiency of inter prediction is further improved, is demand very urgent in Video coding.
At present, when carrying out inter prediction, reconstructed multiple frames can be selected as reference frame.Using multi-reference frame energy Enough video coding performance is brought to be lifted, reason mainly there are two aspects：One is due to the influence of the factors such as camera noise, multi-reference frame And combinations thereof more signal forms can be provided, present encoding block is found more preferable match block；Second, work as video sequence In the presence of when blocking and reappearing in row, present encoding block has the bigger block that may find matching in multi-reference frame.General is newest Video encoding standard High Efficiency Video Coding (HEVC) multi-reference frame is by one closest to frame and several The secondary top-quality frames composition closed on, the management of multi-reference frame are based on reference chart image set (Reference Picture Set, RPS). The selection and management of reference frame are that very the part of core, the influence to overall video coding efficiency are also compared in inter-frame prediction techniques It is larger.
There is a few thing to propose other multi-reference frames selections and management method, such as：
Adaptive reference frame based on video content choose (A.S.Dias, S.Schwarz, M.Siekmann, S.Bosse, H.Schwarz,D.Marpe,J.Zubrzycki and M.Mrak,“Perceptually optimised video compression.”in IEEE International Conference On Multimedia and Expo (ICME2015),Torino,2015.)
Based on improvement reference frame list list sorting (S.Schwarz the and M.Mrak, " Improved for being referenced distribution of content reference picture list sorting in video coding.”in IEEE International Conference on Systems,Signals and Image Processing(IWSSIP 2015),London,2015.)
The shortcomings that above method, is as follows：
1st, the multi-reference frame in HEVC is chosen and management uses same processing scheme to different video contents, without certainly Adaptability.
2nd, the adaptive reference frame based on video content, which is chosen, needs to carry out preanalysis to video to be encoded, it is difficult to applies In the scene of streaming coding.
3rd, had a great influence based on the improvement reference frame list list sorting for being referenced distribution of content by video content to be encoded, and Need by extra reference frame internal memory aid in treatment.
In addition to disadvantages mentioned above, existing reference content management and prioritization scheme are all based on reference frame.In decoding end In the case that reference buffer area (Decoded Picture Buffer, DPB) capacity is limited, content repeats in multi-reference frame, deposits It is not high to the utilization ratio of buffering area in redundancy, limit the further performance boost of Video coding.
The content of the invention
It is an object of the invention to provide a kind of video interprediction encoding method based on reference plate, by optimizing in reference Hold, particularly remove the redundancy between multi-reference frame, make full use of decoding buffer zone capacity, effectively improve the compression of Video coding Efficiency.
The purpose of the present invention is achieved through the following technical solutions：
A kind of video interprediction encoding method based on reference plate, including：
Reference plate generation and management comprising single content：When current reference frame will remove short-term reference frame list, base In the distribution situation of intra-frame prediction block on current reference frame, select the reference plate comprising single content and be added to reference plate buffering Area；If reference plate buffering area is full, what the reference plate priority calculation according to predefined determined to select includes single Whether the reference plate of one content is added to reference plate buffering area；
Reference plate generation and management comprising duplicate contents：Detect static, dynamic present in video sequence and interval weight Multiple content, high quality coding, then point based on high quality encoding block are then carried out to the block of duplicate contents according to testing result Cloth situation generates the reference plate for including duplicate contents；Reference plate comprising duplicate contents is directly added into reference plate buffering area, then Relatively it is managed according to reference plate priority ratio；
Reference plate is retrieved and utilized：For current CTU to be encoded, from reference plate buffering area by the way of Histogram Matching Middle roughing goes out the reference plate of multiple candidates, then a reference plate is chosen by the way of fast motion estimation；The ginseng that will be selected Examine after piece handled by the way of predetermined, with original reference frame in Video coding side by side for the encoding block in current CTU to be encoded The flag bit of reference plate whether will be with reference to during with reference to, coding for indicating the encoding block in current CTU to be encoded, and if by With reference to when, reference plate is incorporated into code stream in the index of reference plate buffering area.
It is described current reference frame will remove short-term reference frame list when, based on current reference frame intra-frame prediction block point Cloth situation, selecting the step of reference plate comprising single content is added to reference plate buffering area includes：
When current reference frame will remove short-term reference frame list, according to specified reference plate size, sliding window is constructed From top to bottom, current reference frame is from left to right scanned, the area for finding out the block of infra-frame prediction in window reaches the position of maximum； If the maximum exceedes the threshold value specified, the reconstruction image regional structure one in corresponding window includes single content Reference plate；The pixel value in reconstruction image region is not only recorded in the reference plate of acquisition, and also includes following information：The position of window Put, the average value of all pieces of quantization step, current reference frame order in the video sequence, reference plate are referenced in reference plate Number and reference plate order；
After selecting a reference plate comprising single content, the block of the infra-frame prediction in the window of its present position is marked For inter prediction, the scanning process of the repetition above, untill it can not select the satisfactory reference plate comprising single content.
If the reference plate buffering area is full, the reference plate priority calculation according to predefined determines to select The reference plate comprising single content whether be added to reference plate buffering area and include：
A reference plate for including single content is often generated, is just added it in reference plate buffering area, if reference plate Buffering area is full；Then the reference plate priority calculation according to predefined calculates reference plate and choosing in reference plate buffering area The priority of the reference plate comprising single content gone out, calculation formula are：
Pc=(QP0-QPc)×NQP+Ic+Trc×Nr；
In above formula, PcFor the priority of reference plate, QP0Be setting some constant, QPcIt is the average qp value of reference plate, NQP With NrIt is for specified constant, IcFor the order of reference plate, TrcThe number of block reference is encoded for reference plate, is specified constant；
Join if the priority for the reference plate comprising single content selected is more than lowest priority in reference plate buffering area The priority of piece is examined, then the reference of lowest priority in reference plate buffering area is replaced with the reference plate comprising single content selected Piece；Otherwise, the reference plate for including single content selected is abandoned.
The content that static present in the detection video sequence, dynamic and interval repeat includes：
The detection unit of different duplicate contents is code tree unit CTU, after the CTU of duplicate contents is detected, prediction The number of repetition of duplicate contents in corresponding CTU；Detailed process is as follows：
Static duplicate contents detection：The average of each CTU internal variances in image is calculated, if current CTU average is less than The threshold value of setting, mark current CTU to be repeated to be static, and estimate the number of repetition L of static duplicate contentsA=S × frame per second, herein Afterwards in the S seconds, detection dynamic and the content of spaced repetition not at the CTU comprising static duplicate contents；
Dynamic duplicate contents detect：For the CTU of non-static repetition, with original of the CTU original pixels on its reference frame Beginning pixel is taken exercises estimation, obtain corresponding to motion vector and corresponding estimation residual error；It is all small if all of residual values In the threshold value of setting, infer dynamic duplicate contents in corresponding CTU be present, and assume dynamic duplicate contents within a period of time Even motion, moving averages MVmeanFor：
Wherein, MViThe motion vector estimated for CTU and its some reference frame, diFrame and ginseng where corresponding CTU The time interval of frame is examined, N is the number of reference frame；After obtaining moving averages, then estimate the number of repetition L of dynamic duplicate contentsB：
LB=min { Lx,Ly}；
Wherein, Lx、LyThe existence time that respectively horizontal, vertical direction is estimated, W, H are respectively the width of video sequence, height, Cx、CyRespectively current CTU horizontal stroke, ordinate, MVXmean、MVYmeanFor MVmeanX, y-component；Detect and repeated comprising dynamic After the CTU of content, in estimating in the range of existence time for dynamic duplicate contents, new dynamic duplicate contents are no longer detected；
It is spaced duplicate contents detection：When current reference piece is referred to by some CTU, then corresponding CTU is on current reference piece Total cost of fast motion estimation will be recorded；Meanwhile the position that is matched in fast motion estimation of current reference piece also by Record；If current reference piece is repeatedly referred to, current reference piece need retain the total cost of fast motion estimation average and The union of matched position；When being encoded to M frames, for each CTU on M frames, reference plate is carried out in reference plate buffering area Retrieval, if the total cost of obtained fast motion estimation is less than the total cost average of fast motion estimation of current reference piece record Certain multiple, and the position matched includes the certain multiple of the matched position union of current reference piece record, then judges corresponding CTU content exists with the content in the reference plate retrieved to be repeated；If the reference plate retrieved is drawn before M frames are compiled With l times, infer in follow-up cataloged procedure, similar duplicate contents there is also LCIt is secondary, LCCalculation such as following formula institute Show：
Wherein, P is the length of video sequence.
It is described that the block progress high quality coding of duplicate contents is included according to testing result：
It will be repeated L times assuming that some duplicate contents is estimated, then calculate optimum encoding parameter for corresponding duplicate contents, compiled The determination of code parameter also determines in CTU levels, is shown below：
Wherein, λ is the basic LaGrange parameter of whole video sequence, λiIt is CTU iths corresponding to some duplicate contents The coding parameter set during appearance；Coding parameter λ is determinediAfterwards, corresponding QPiObtained according to following formula：
QPi=4.2005 × ln (λi)+13.7122
The QP obtainediUsing as quantization step during corresponding CTU codings；Meanwhile for using coding parameter λ1CTU, mark Remember that the CTU encodes for high quality, and its QP value is incorporated into code stream.
The distribution situation generation based on high quality encoding block includes the reference plate of duplicate contents；Include duplicate contents Reference plate is directly added into reference plate buffering area, then according to reference plate priority ratio be relatively managed including：
Distribution situation generation based on high quality coding CTU includes the reference plate of duplicate contents, and its step is as follows：Encode After present frame, according to specified reference plate size, construction sliding window from top to bottom, from left to right scan current reference frame, look for The intraoral CTU quantity for carrying out high quality coding of exit window reaches the position of maximum；If the maximum exceedes the threshold value specified, Reference plate is generated with the pixel value in window；The reference plate of generation is the reference plate for including duplicate contents, wherein also including such as Lower information：All CTU average qp value in pixel value, window；After the reference plate generation of duplicate contents, by the height of its present position Quality coded mark is changed to regular quality coding, continuing the scanning process of the above, duplicate contents can be generated until can not select Reference plate untill；
The reference plate comprising duplicate contents of generation is directly added into reference plate buffer area；Mode is as follows：If include repetition In the reference plate of content, there is the reference plate that one or more CTU with reference to, then replaced with the reference plate comprising duplicate contents Change in reference plate buffering area that an existing reference plate, the reference plate being specifically replaced are currently to include duplicate contents All CTU choose that most reference plate of reference number of times in reference plate；Otherwise, ginseng is replaced with the reference plate comprising duplicate contents Examine the reference plate that priority is minimum in piece buffering area；
Afterwards, the reference plate comprising duplicate contents of reference plate buffering area is relatively managed according to reference plate priority ratio.
Described roughing goes out the reference plate of multiple candidates from reference plate buffering area by the way of Histogram Matching, then uses The mode of fast motion estimation, which chooses the step of reference plate, to be included：
Histogram Matching：Calculate first straight between all reference plates in current CTU to be encoded and reference plate buffering area Square map distance, then result of the less reference plate of multiple Histogram distances as Histogram Matching is selected using predetermined way；
The calculation of Histogram distance is as follows：1) its histogram is calculated to current CTU to be encoded：Will be current to be encoded CTU luminance component is divided into the block of 4 32 × 32, the histogram of one H dimension of every piece of statistics, to two chromatic components of U and V, The histogram of each one H dimension of statistics；Six H are tieed up into histogram and are spliced to form 6H dimension histograms；2) to every in reference plate buffering area The block of 64 × 64 sizes in one reference plate, with level, several pixels are once sampled with vertical direction interval, to sampling Block calculate histogram；3) it is right one by one to carry out the histogram of current CTU to be encoded histogram and all sampling blocks of reference plate Than calculating the distance between histogram, and select wherein minimum distance as current CTU to be encoded and the histogram of reference plate Distance；
Fast motion estimation：In the multiple reference plates obtained from Histogram Matching, the method for fast motion estimation, choosing are used Go out a reference plate most possibly utilized；Current CTU to be encoded carries out the calculating side of fast motion estimation with each reference plate Formula is as follows：Current CTU is divided into the fritter that fixed size is R × R, each R × R block search cost in current reference piece Minimum block, wherein cost are made up of the encoder bit rate of absolute difference, sad value and motion vector；The mode of search is first by every five The raster scanning of point determines the starting point searched in reference plate, then around the starting point up and down in certain pixel coverage Carry out 8 brilliant search；The cost of all R × R block is added up as current CTU to be encoded on current reference piece Overall cost, and then filter out corresponding reference plate during overall Least-cost.
It is described to use predetermined mode to handle for new reference frame the reference plate selected, and by itself and Video coding Central Plains Having reference frame to be referred to side by side for current block to be encoded includes：
The reference plate selected is positioned over blank with reference frame formed objects according to position when being chosen from reference frame On；Meanwhile by the reference plate for being in same frame in reference plate buffering area with the reference plate selected also together be placed on blank, then Using the blank reference frame new as one；
Original reference frame in new reference frame and Video coding is arranged side by side, used when being encoded for current block to be encoded；Wherein, The starting point that current block to be encoded is searched for when carrying out inter prediction encoding on new reference frame, during estimation is set to histogram With the optimum position found；Motion-vector prediction MVP choosing method is as follows：Check five positions around current prediction unit PU Whether the block put employs reference plate, and and if only if uses reference plate, and its reference plate is come with the reference plate used in current PU From same reference frame when, then the motion vector MV on relevant position is as candidate MVP；Two MVP are therefrom chosen in order to add Enter candidate's MVP lists, if the available MVP obtained on locus less than two, is used where current block to be encoded successively Position point to the MV of reference plate center and (0,0) is filled, untill filling up two.
The flag bit that reference plate whether will be with reference to for indicating the encoding block in current CTU to be encoded, Yi Jiruo With reference to when, reference plate is incorporated into code stream in the index of reference plate buffering area to be included：
Whether with reference to reference plate flag bit be one 0 or 1 flag bit, flag bit two entered using context-adaptive Whether arithmetic coding processed, the context model used have three, by the CTU on current the CTU left sides and top to be encoded with reference to reference Piece determines, if two CTU are predicted without using reference plate, with first context model；If one of them is employed Reference plate is predicted, then with second context model；If both having used reference plate, with the 3rd context model；
If current CTU to be encoded with reference to reference plate, reference plate uses block code, coding length in the index of buffering area Spend the reference plate number depending on reference plate buffering area when reference plate is referenced；
Afterwards, by corresponding flag bit, code stream is incorporated into index of the reference plate in buffering area.
This method also includes：Whether current CTU to be encoded has carried out high quality coding, and if carry out high quality coding, QP when then high quality encodes is relative to normal QP adjusted value, it is necessary to be incorporated into code stream；It is specific as follows：
Whether carry out the flag bit of high quality coding be one 0 or 1 flag bit, flag bit is using context-adaptive Whether binary arithmetic coding, the context model used have three, carried out by the CTU on current the CTU left sides and top to be encoded High quality is encoded to determine, if two CTU are no to carry out high quality coding, with first context model；If wherein one Individual progress high quality coding, then with second context model；If high quality coding is both carried out, with the 3rd context Model；
If carrying out high quality coding, QP when high quality encodes uses block code relative to normal QP adjusted value.
As seen from the above technical solution provided by the invention, the reference content management based on reference plate and optimization side Case, more information can be collected under conditions of DPB capacity is limited so that present encoding block can more maximum probability find With block.Finally, overall effect is in the case where being slightly increased coding and decoding end complexity, effectively improves Video coding Efficiency.
Brief description of the drawings
In order to illustrate the technical solution of the embodiments of the present invention more clearly, required use in being described below to embodiment Accompanying drawing be briefly described, it should be apparent that, drawings in the following description are only some embodiments of the present invention, for this For the those of ordinary skill in field, on the premise of not paying creative work, other can also be obtained according to these accompanying drawings Accompanying drawing.
Fig. 1 is a kind of flow chart of the video interprediction encoding method based on reference plate provided in an embodiment of the present invention；
Fig. 2 is the flow chart of reference plate of the generation comprising single content provided in an embodiment of the present invention；
Fig. 3 is the flow chart of reference plate of the generation comprising duplicate contents provided in an embodiment of the present invention；
Fig. 4 is that reference plate provided in an embodiment of the present invention is retrieved and the flow chart using process；
Fig. 5 is the schematic diagram provided in an embodiment of the present invention for calculating current CTU histograms to be encoded.
Embodiment
With reference to the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete Ground describes, it is clear that described embodiment is only part of the embodiment of the present invention, rather than whole embodiments.Based on this The embodiment of invention, the every other implementation that those of ordinary skill in the art are obtained under the premise of creative work is not made Example, belongs to protection scope of the present invention.
Fig. 1 is a kind of flow chart of the video interprediction encoding method based on reference plate provided in an embodiment of the present invention. As shown in figure 1, it mainly comprises the following steps：
Step 1, reference plate generation and management comprising single content：Short-term reference frame list will be removed in current reference frame When, based on the distribution situation of intra-frame prediction block on current reference frame, select the reference plate comprising single content and be added to reference plate Buffering area；If reference plate buffering area is full, the reference plate priority calculation according to predefined determines the bag selected Whether the reference plate containing single content is added to reference plate buffering area；
Step 2, reference plate generation and management comprising duplicate contents：Detect static, dynamic present in video sequence and The content repeated is spaced, high quality coding is then carried out to the block of duplicate contents according to testing result, then encode based on high quality The distribution situation generation of block includes the reference plate of duplicate contents；Reference plate comprising duplicate contents is directly added into reference plate buffering Area, then relatively it is managed according to reference plate priority ratio；
Step 3, reference plate retrieval and utilization：For current CTU (code tree unit) to be encoded, using Histogram Matching Mode roughing from reference plate buffering area goes out the reference plate of multiple candidates, then one is chosen by the way of fast motion estimation Reference plate；Use predetermined mode to handle for new reference frame the reference plate selected, and by its with it is original in Video coding Reference frame is side by side for the encoding block reference in current CTU to be encoded, by for indicating the coding in current CTU to be encoded during coding Whether block with reference to the flag bit of reference plate, and if when being referenced, reference plate is incorporated into code stream in the index of reference plate buffering area.
In order to make it easy to understand, elaborated below for these three processes.
First, reference plate generation and management comprising single content.
The process mainly includes：Generation includes the reference plate of single content；Management includes the reference plate of single content；Specifically It is as follows：
1st, generation includes the reference plate of single content.
As shown in Fig. 2 reference plate process of the generation comprising single content is as follows：Short term reference will be removed in current reference frame During frame list, according to specified reference plate size, construction sliding window from top to bottom, from left to right scan current reference frame, look for The area of the block of the intraoral infra-frame prediction of exit window reaches the position of maximum；If the maximum exceedes the threshold value specified, right Answer one reference plate for including single content of reconstruction image regional structure in window；The reference plate of acquisition not only records reconstruction figure As the pixel value in region, in addition to following information：All pieces of quantization step in the position of window, reference plate Average value, the order (Picture of current reference frame in the video sequence of (Quantization Parameter, QP) Order Count, POC), the order of the number that reference plate is referenced and reference plate；
After selecting a reference plate comprising single content, the block of the infra-frame prediction in the window of its present position is marked For inter prediction, the scanning process of the repetition above, untill it can not select the satisfactory reference plate comprising single content.
It is exemplary, for the allocation plan of reference plate number of the current reference frame under different resolution image and size As shown in table 1.
Image resolution ratio Reference plate size Reference plate number
416*240 80*80 16
832*480 112*112 32
1024*768 112*112 64
1280*720 120*120 64
1920*1080 256*256 64
The allocation plan of reference plate number and size under the different resolution image of table 1
2nd, management includes the reference plate of single content.
A reference plate for including single content is often generated, is just added it in reference plate buffering area.
If there is clear position in reference plate buffering area, can directly by a upper process select comprising single content Reference plate is directly placed into.
If reference plate buffering area is full, need to calculate reference according to the reference plate priority calculation of predefined The priority of reference plate and the reference plate comprising single content selected in piece buffering area, calculation formula are：
Pc=(QP0-QPc)×NQP+Ic+Trc×Nr；
In above formula, PcFor the priority of reference plate, QP0Be setting some constant, QPcIt is the average qp value of reference plate, NQP With NrIt is specified constant, IcFor the order of reference plate, TrcThe number of block reference is encoded for reference plate.It should be noted that When reference plate is put into buffering area, its order IcIt is that the reference plate number having had in current buffer adds 1 (i.e. first puts The reference plate order for entering buffering area is 1, by that analogy)；If reference plate is selected reference in an encoding process, it The order that order can be updated to order maximum reference plate in now buffering area adds 1.In addition, when reference plate is newly put into buffering area, It is referenced number Trc0 is set to, hereafter, when reference plate is selected reference in an encoding process, it is referenced number pair 1 should be increased.
And then the reference plate by lowest priority in comparison reference piece buffering area and the new ginseng for including single content The relative height of priority of piece is examined, to decide whether the new reference plate comprising single content being put into reference plate buffering area.Such as The priority for the reference plate comprising single content that fruit is selected be more than reference plate buffering area in lowest priority reference plate it is preferential Level, then the reference plate of lowest priority in reference plate buffering area is replaced with the reference plate comprising single content selected；Otherwise, lose Abandon the reference plate for including single content selected.
2nd, reference plate generation and management comprising duplicate contents
The process mainly includes：Generation includes the reference plate of duplicate contents；Management includes the reference plate of duplicate contents；Specifically It is as follows：
1st, generation includes the reference plate of duplicate contents.
Generally, the content largely repeated in video sequence be present, the present invention proposes to examine the duplicate contents in video sequence Measure and, carry out high quality coding and be stored in reference plate to refer to for next code, the reference plate generation comprising duplicate contents Flow chart is as shown in figure 3, mainly include three parts：1) duplicate contents detect；2) coding parameter determines；3) generation includes repetition The reference plate of content；The detailed process of these three parts is as follows：
1) duplicate contents detect.
In the embodiment of the present invention, the duplicate contents in video sequence are divided into three classes：Static duplicate contents, dynamically repeat in Hold, interval duplicate contents.It is discussed below, for the duplicate contents of different situations, used corresponding detection scheme.At these Detection unit employed in detection scheme is the code tree unit (coding tree unit, CTU) in HEVC, is being detected After duplicate contents CTU, the number of repetition of the CTU can be also predicted.
A, static duplicate contents detection：Static duplicate contents in video sequence, such as the background in meeting sequence, often Fixed space position in video repeats a period of time.During coding, original pixels and its ginseng based on present frame The variance of each pixel position can be obtained by examining the original pixels of frame.With the progress of coding, multigroup continuous side is collected Difference, and the foundation for detecting static duplicate contents is used as by the use of the poor average of same position pixel prescription.Calculate in image in each CTU The average of variance, if current CTU average is less than the threshold value of setting, mark current CTU to be repeated to be static, and estimate static weight The number of repetition L of multiple contentA=S × frame per second, thereafter in the S seconds, not at the CTU comprising static duplicate contents detection dynamic and The content of spaced repetition.Exemplary, S=10 can be set herein, that is, assume static repeat 10 seconds.
B, dynamic duplicate contents detect：Dynamic duplicate contents in video sequence, the back of the body changed such as camera lens is mobile Scape and the personage that moves in video etc., the mode based on estimation detects.For the CTU of non-static repetition, with CTU original Original pixels of the beginning pixel on its reference frame are taken exercises estimation, obtain corresponding to motion vector and corresponding estimation it is residual Difference；The threshold value set is both less than if all of residual values, infers dynamic duplicate contents in corresponding CTU be present, and is assumed dynamic State duplicate contents Uniform Movement within a period of time, moving averages MVmeanFor：
Wherein, MViThe motion vector estimated for CTU and its some reference frame, diFrame and ginseng where corresponding CTU The time interval of frame is examined, N is the number of reference frame；After obtaining moving averages, then estimate the number of repetition L of dynamic duplicate contentsB：
LB=min { Lx,Ly}；
Wherein, Lx、LyThe existence time that respectively horizontal, vertical direction is estimated, W, H are respectively the width of video sequence, height, Cx、CyRespectively current CTU horizontal stroke, ordinate, MVXmean、MVYmeanFor MVmeanX, y-component；Detect and repeated comprising dynamic After the CTU of content, in estimating in the range of existence time for dynamic duplicate contents, new dynamic duplicate contents are no longer detected；
C, it is spaced duplicate contents detection：Persistently occur in the video sequence not with static duplicate contents and dynamic duplicate contents Together, spaced repetition content typically can not find matching in the range of its short-term reference frame, between the present invention is detected based on reference plate Every the content that property repeats.When current reference piece is referred to by some CTU, then quick motions of the corresponding CTU on current reference piece Total cost of estimation will be recorded (fast motion estimation process will be introduced later)；Meanwhile current reference piece is quick The position matched during estimation is also recorded；If current reference piece is repeatedly referred to, current reference piece needs to retain soon The average of the fast total cost of estimation and the union of matched position；When being encoded to M frames, for each CTU on M frames, Reference plate retrieval is carried out in reference plate buffering area, if the total cost of obtained fast motion estimation is less than current reference piece record The certain multiple of the total cost average of fast motion estimation, and the position matched includes the matched position of current reference piece record simultaneously The certain multiple of collection, then judge that corresponding CTU content exists with the content in the reference plate retrieved and repeat；If retrieve Reference plate has been cited l times before M frames are compiled, and infers in follow-up cataloged procedure, and similar duplicate contents there is also LCIt is secondary, LCCalculation be shown below：
Wherein, P is the length of video sequence.
2) coding parameter determines.
It will be repeated L times assuming that some duplicate contents is estimated, then calculate optimum encoding parameter for corresponding duplicate contents, compiled The determination of code parameter also determines in CTU levels, is shown below：
Wherein, λ is the basic LaGrange parameter of whole video sequence, λiIt is CTU iths corresponding to some duplicate contents The coding parameter set during appearance；From the point of view of first formula above, λ1CTU the 1st time namely corresponding to some duplicate contents The coding parameter set during appearance is relatively low, and just uses second formula since the 2nd time, that is to say equal to normal parameter.
Coding parameter λ is determinediAfterwards, corresponding QPiObtained according to following formula：
QPi=4.2005 × ln (λi)+13.7122
Afterwards, the QP that will can be obtainediUsing as quantization step during corresponding CTU codings；Meanwhile for using coding ginseng Number λ1CTU, mark the CTU to be encoded for high quality, and its QP value be incorporated into code stream.
3) generation includes the reference plate of duplicate contents.
In the embodiment of the present invention, the distribution situation generation based on high quality coding CTU includes the reference plate of duplicate contents, its Generating process is similar with the reference plate comprising single content explained before；It is specific as follows：
After having encoded present frame, according to specified reference plate size, construction sliding window from top to bottom, from left to right scans Current reference frame, the CTU quantity for finding out progress high quality coding in window reach the position of maximum；If the maximum exceedes The threshold value specified, then generate reference plate with the pixel value in window；The reference plate of generation is the reference plate for including duplicate contents, Including information as the information of single content reference plate, such as in pixel value, window all CTU average qp value Deng；After the reference plate generation of duplicate contents, the high quality code identification of its present position is changed to regular quality coding, repeated More than scanning process, untill it can not select and can generate the reference plate of duplicate contents；
The reference plate comprising duplicate contents of generation is directly added into reference plate buffer area.Mode is as follows：If include repetition In the reference plate of content, there is the reference plate that one or more CTU with reference to, then replaced with the reference plate comprising duplicate contents Change in reference plate buffering area that an existing reference plate, the reference plate being specifically replaced are currently to include duplicate contents All CTU choose that most reference plate of reference number of times in reference plate；Otherwise, ginseng is replaced with the reference plate comprising duplicate contents Examine the reference plate that priority is minimum in piece buffering area.
2nd, management includes the reference plate of duplicate contents.
The reference plate for including duplicate contents of reference plate buffering area, according to the calculating side of reference plate priority described previously Formula, its priority is calculated, and same with single content reference plate be managed.
3rd, reference plate retrieval and utilization.
When several existing reference plates in reference plate buffering area, some block is encoded, it is possible to carried out using reference plate Inter prediction.Reference plate retrieval is with the flow chart using process as shown in figure 4, being mainly made up of three parts：Reference plate retrieval, Coding selection rate-distortion optimization (Rate Distortion Optimization, RDO) and the syntax knot for reference plate Structure, introduce separately below.
1st, reference plate is retrieved
Because the reference plate quantity of reference plate buffering area may be more, coding side can be dramatically increased by carrying out estimation one by one Complexity, so in the embodiment of the present invention, retrieve one using the method for Histogram Matching and fast motion estimation first The reference plate most possibly utilized, further estimation is then done in the reference plate retrieved.Histogram Matching and quick The specific practice of estimation is as follows.
Histogram Matching：Calculate first straight between all reference plates in current CTU to be encoded and reference plate buffering area Square map distance, then result of the less reference plate of multiple Histogram distances as Histogram Matching is selected using predetermined way.Show Example property, after calculating all Histogram distances, it can be sorted successively according to descending order, then end is selected from sequence Result of multiple Histogram distances as Histogram Matching.
The calculation of Histogram distance is as follows：1) its histogram is calculated to current CTU to be encoded：Will be current to be encoded CTU luminance component is divided into the block of 4 32 × 32, the histogram of one H dimension of every piece of statistics, to two chromatic components of U and V, The histogram of each one H dimension of statistics；Six H are tieed up into histogram and are spliced to form 6H dimension histograms, said process can be found in Fig. 5；2) To the block of 64 × 64 sizes in each reference plate in reference plate buffering area, with level, several pixels are entered with vertical direction interval Row is once sampled, and histogram is calculated to the block of sampling；3) by current CTU to be encoded histogram and all sampling blocks of reference plate Histogram is contrasted one by one, calculates the distance between histogram, and selects wherein minimum distance as current CTU to be encoded With the Histogram distance of reference plate.Exemplary, above-mentioned H can be set to 32.
Fast motion estimation：In the multiple reference plates obtained from Histogram Matching, the method for fast motion estimation, choosing are used Go out a reference plate most possibly utilized；Current CTU to be encoded carries out the calculating side of fast motion estimation with each reference plate Formula is as follows：Current CTU is divided into the fritter that fixed size is R × R, each R × R block search cost in current reference piece Minimum block, wherein cost are by absolute difference and (Sum of Absolute Difference, SAD) value and the coding of motion vector Code check forms；The mode of search is the first starting point by determining to search in reference plate every 5 points of raster scanning, is then risen at this Around initial point up and down in certain pixel coverage (such as：In 64 pixel coverages) carry out 8 brilliant search；By all R × The cost of R block is cumulative as current overall costs of the CTU to be encoded on current reference piece, and then filters out overall cost most A reference plate corresponding to hour.Meanwhile the foundation that this cost will be also recorded as interval duplicate contents detection.Example Property, above-mentioned R can be set to 16.
2nd, coding selection RDO
After retrieving reference plate, by the reference plate selected according to from reference frame choose when locus be positioned over On the blank of reference frame formed objects；Meanwhile the reference plate that same frame will be in reference plate buffering area with the reference plate selected Also it is placed on together on blank, then using the blank reference frame new as one, it is arranged side by side with original reference frame in Video coding, Used when being encoded for the block in current CTU to be encoded.
Current block to be encoded carries out inter prediction encoding in Video coding on original reference frame when, using original side Formula.And when carrying out inter prediction encoding on new reference frame, there is following difference with the mode of original reference frame：Estimation When the starting point searched for be set to the optimum position that Histogram Matching described previously is found；Motion-vector prediction MVP choosing method It is as follows：Check whether the block around current prediction unit (prediction unit, PU) on five positions employs reference plate (the either duplicate contents of single content), and if only if uses reference plate, and used in its reference plate and current PU When reference plate comes from same reference frame, then the motion vector MV on relevant position is as candidate MVP；Therefrom press original mode The order specified chooses two MVP and adds candidate's MVP lists, if the available MVP obtained on locus less than two, successively Filled using MV and (0,0) that reference plate center is pointed to from the position of the CTU where current block to be encoded, filling up two is Only.
3rd, for the syntactic structure of reference plate
This process is mainly related to coding；Whether some CTU to be encoded with reference to new reference frame, if with reference to, Then think that CTU to be encoded with reference to reference plate corresponding to new reference frame, during coding reference plate reference plate buffering area index, It is required for being incorporated into code stream.
First, if the flag bit with reference to reference plate be one 0 or 1 flag bit, flag bit uses context-adaptive Binary arithmetic coding (Context Adaptive Binary Arithmetic Coding, CABAC), use up and down Literary model has three, and whether reference plate is referred to determine, if two CTU do not have by the CTU on current the CTU left sides and top to be encoded Have and predicted using reference plate, then with first context model；If one of them employs reference plate prediction, with second Hereafter model；If both having used reference plate, with the 3rd context model；If current CTU to be encoded with reference to reference plate, Then reference plate uses block code in the index of buffering area, and code length is depending on reference plate buffering area when reference plate is referenced Reference plate number；For example, with 3 code bits (bits) during 5-8 reference plate, with 4 bits during 9-16 reference plate, with such Push away.Afterwards, by corresponding flag bit, code stream is incorporated into index of the reference plate in buffering area.
Encoded it will be understood by those skilled in the art that a CTU can be decomposed into multiple pieces, Histogram Matching and fast Fast estimation is done to CTU, and a reference plate can be found out to each CTU.Then, in the more of the subordinate to this CTU When individual encoding block carries out coding selection RDO, original reference frame in Video coding can be used, reference plate can also be utilized to generate New reference frame.Finally, after this CTU all encoding blocks all finish RDO, actual coding is carried out, if this There is at least one encoding block to select new reference frame in CTU, then it is assumed that this CTU with reference to reference plate, compile a flag bit 1, and encode the index for choosing reference plate；Otherwise, illustrate that this CTU does not select reference plate, compile a flag bit 0.
In addition, whether current CTU to be encoded has carried out high quality coding, and if carrying out high quality coding, high quality QP during coding is relative to normal QP adjusted value, it is necessary to be incorporated into code stream；It is specific as follows：Whether the mark of high quality coding is carried out Position be one 0 or 1 flag bit, flag bit use context-adaptive binary arithmetic coding, the context model used There are three, determined by whether the CTU on current the CTU left sides and top to be encoded carries out high quality coding, if two CTU do not have High quality coding is carried out, then with first context model；If one of carry out high quality coding, with second context Model；If high quality coding is both carried out, with the 3rd context model；If carrying out high quality coding, high quality is compiled QP during code uses block code relative to normal QP adjusted value.
It will be understood by those skilled in the art that the context model being previously mentioned refers to that the probability for arithmetic coding is pre- Model is surveyed, it is similar with context model general in Video coding.
On the other hand, in order to illustrate the effect of such scheme, also it is tested based on such scheme.Test condition bag Include：(low delay B is Low-delay B, LDB for interframe configuration；Low delay P is Low-delay P, LDP), basic quantization step-length (QP) { 22,27,32,37 } are arranged to, based on software be HM16.7, cycle tests is the classes of B-F five in the logical sequencing row of HEVC. Baseline is set to：HM16.7 uses 5 reference frames, and contrast scheme is set to HM using 4 reference frames plus being equivalent to 1 reference frame institute With several reference plates of internal memory, wherein lower 1 reference frame of different resolution reference plate number of equal value and size is shown in and can joined The table 1 seen above.Experimental result see the table below 2~3, and wherein table 2 is the performance comparison under LDB, LDP are set, and table 3 is LDB, LDP Encoding and decoding end complexity contrast under setting.As can be seen that suggesting plans can distinguish relative to baseline under LDB and LDP patterns The code check for obtaining 5.1% and 5.0% is saved, and the complexity at encoding and decoding end is not significantly increased.
Performance comparison under table 2 LDB, LDP setting
Mode LDB LDP
Encoding Time 129% 142%
Decoding Time 124% 123%
Encoding and decoding end complexity under table 3 LDB, LDP are set contrasts
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment can To be realized by software, the mode of necessary general hardware platform can also be added by software to realize.Based on such understanding, The technical scheme of above-described embodiment can be embodied in the form of software product, the software product can be stored in one it is non-easily In the property lost storage medium (can be CD-ROM, USB flash disk, mobile hard disk etc.), including some instructions are causing a computer to set Standby (can be personal computer, server, or network equipment etc.) performs the method described in each embodiment of the present invention.
The foregoing is only a preferred embodiment of the present invention, but protection scope of the present invention be not limited thereto, Any one skilled in the art is in the technical scope of present disclosure, the change or replacement that can readily occur in, It should all be included within the scope of the present invention.Therefore, protection scope of the present invention should be with the protection model of claims Enclose and be defined.

Claims (8)

  1. A kind of 1. video interprediction encoding method based on reference plate, it is characterised in that including：
    Reference plate generation and management comprising single content：Current reference frame will remove short-term reference frame list when, based on work as The distribution situation of intra-frame prediction block on preceding reference frame, select the reference plate comprising single content and be added to reference plate buffering area；Such as Fruit reference plate buffering area is full, then the reference plate priority calculation according to predefined determines that selects includes single content Reference plate whether be added to reference plate buffering area；
    Reference plate generation and management comprising duplicate contents：Detect what static, dynamic and interval present in video sequence repeated Content, high quality coding, then the distribution feelings based on high quality encoding block are then carried out to the block of duplicate contents according to testing result Condition generates the reference plate for including duplicate contents；Reference plate comprising duplicate contents is directly added into reference plate buffering area, then according to Reference plate priority ratio is relatively managed；
    Reference plate is retrieved and utilized：It is thick from reference plate buffering area by the way of Histogram Matching for current CTU to be encoded The reference plate of multiple candidates is selected, then a reference plate is chosen by the way of fast motion estimation；The reference that will be chosen After piece is handled by the way of predetermined, join side by side for the encoding block in current CTU to be encoded with original reference frame in Video coding Examine, whether will with reference to the flag bit for the reference plate chosen during coding for indicating the encoding block in current CTU to be encoded, with And if when being referenced, the reference plate chosen is incorporated into code stream in the index of reference plate buffering area；
    If the reference plate buffering area is full, the reference plate priority calculation according to predefined determines the bag selected Whether the reference plate containing single content, which is added to reference plate buffering area, includes：
    A reference plate for including single content is often generated, is just added it in reference plate buffering area, if reference plate buffers Area is full；Then the reference plate priority calculation according to predefined calculates reference plate in reference plate buffering area and selected The priority of reference plate comprising single content, calculation formula are：
    Pc=(QP0-QPc)×NQP+Ic+Trc×Nr；
    In above formula, PcFor the priority of reference plate, QP0Be setting some constant, QPcIt is the average qp value of reference plate, NQPWith Nr It is for specified constant, IcFor the order of reference plate, TrcThe number of block reference is encoded for reference plate, is specified constant；
    If the priority for the reference plate comprising single content selected is more than lowest priority reference plate in reference plate buffering area Priority, then with the reference plate of lowest priority in the reference plate replacement reference plate buffering area selected comprising single content； Otherwise, the reference plate for including single content selected is abandoned；
    It is described that the block progress high quality coding of duplicate contents is included according to testing result：
    It will be repeated L times assuming that some duplicate contents is estimated, then calculate optimum encoding parameter, coding ginseng for corresponding duplicate contents Several determinations also determines in CTU levels, is shown below：
    Wherein, LAFor the number of repetition of static duplicate contents, LBFor the number of repetition of dynamic duplicate contents, LCFor in the repetition of interval The number of repetition of appearance, λ are the basic LaGrange parameters of whole video sequence, λiIt is CTU iths corresponding to some duplicate contents The coding parameter set during appearance；Coding parameter λ is determinediAfterwards, corresponding QPiObtained according to following formula：
    QPi=4.2005 × ln (λi)+13.7122
    The QP obtainediUsing as quantization step during corresponding CTU codings；Meanwhile for using coding parameter λ1CTU, mark should CTU is encoded for high quality, and its QP value is incorporated into code stream；
    By the reference plate chosen handled by the way of predetermined including：The reference plate chosen is chosen according to from reference frame When position be positioned over on the blank of reference frame formed objects；Meanwhile by reference plate buffering area with the reference plate chosen Reference plate in same frame is also placed on blank together, then using the blank reference frame new as one.
  2. 2. a kind of video interprediction encoding method based on reference plate according to claim 1, it is characterised in that described When current reference frame will remove short-term reference frame list, based on the distribution situation of intra-frame prediction block on current reference frame, select Reference plate comprising single content, which is added to the step of reference plate buffering area, to be included：
    When current reference frame will remove short-term reference frame list, according to specified reference plate size, construction sliding window is from upper Under, current reference frame is from left to right scanned, the area for finding out the block of infra-frame prediction in window reaches the position of maximum；If The maximum exceedes the threshold value specified, then one reference for including single content of reconstruction image regional structure in corresponding window Piece；The pixel value in reconstruction image region is not only recorded in the reference plate of acquisition, and also includes following information：The position of window, ginseng Examine the order in the video sequence of the average value of all pieces of quantization step, current reference frame in piece, time that reference plate is referenced The order of number and reference plate；
    After selecting a reference plate comprising single content, the block of the infra-frame prediction in the window of its present position is labeled as frame Between predict, repeat the above scanning process, untill it can not select the satisfactory reference plate comprising single content.
  3. 3. a kind of video interprediction encoding method based on reference plate according to claim 1, it is characterised in that described The content that static, dynamic and interval repeat present in detection video sequence includes：
    The detection unit of different duplicate contents is code tree unit CTU, and after the CTU of duplicate contents is detected, prediction is corresponding The number of repetition of duplicate contents in CTU；Detailed process is as follows：
    Static duplicate contents detection：The average of each CTU internal variances in image is calculated, if current CTU average is less than setting Threshold value, mark current CTU to be repeated to be static, and estimate the number of repetition L of static duplicate contentsA=S × frame per second, thereafter S In second, detection dynamic and the content of spaced repetition not at the CTU comprising static duplicate contents；
    Dynamic duplicate contents detect：For the CTU of non-static repetition, with original image of the CTU original pixels on its reference frame Element is taken exercises estimation, obtain corresponding to motion vector and corresponding estimation residual error；Both less than set if all of residual values Fixed threshold value, infer dynamic duplicate contents in corresponding CTU be present, and assume that dynamic duplicate contents are uniformly transported within a period of time It is dynamic, moving averages MVmeanFor：
    <mrow> <msub> <mi>MV</mi> <mrow> <mi>m</mi> <mi>e</mi> <mi>a</mi> <mi>n</mi> </mrow> </msub> <mo>=</mo> <mfrac> <mn>1</mn> <mi>N</mi> </mfrac> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>N</mi> </munderover> <mfrac> <mrow> <msub> <mi>MV</mi> <mi>i</mi> </msub> </mrow> <msub> <mi>d</mi> <mi>i</mi> </msub> </mfrac> </mrow>
    Wherein, MViThe motion vector estimated for CTU and its some reference frame, diFrame and reference frame where corresponding CTU Time interval, N be reference frame number；After obtaining moving averages, then estimate the number of repetition L of dynamic duplicate contentsB：
    LB=min { Lx,Ly}；
    <mrow> <msub> <mi>L</mi> <mi>x</mi> </msub> <mo>=</mo> <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mfrac> <mrow> <mi>W</mi> <mo>-</mo> <msub> <mi>C</mi> <mi>x</mi> </msub> </mrow> <mrow> <mo>-</mo> <msub> <mi>MVX</mi> <mrow> <mi>m</mi> <mi>e</mi> <mi>a</mi> <mi>n</mi> </mrow> </msub> </mrow> </mfrac> <mi>M</mi> <mi>V</mi> <msub> <mi>X</mi> <mrow> <mi>m</mi> <mi>e</mi> <mi>a</mi> <mi>n</mi> </mrow> </msub> <mo>&lt;</mo> <mn>0</mn> </mtd> </mtr> <mtr> <mtd> <mfrac> <msub> <mi>C</mi> <mi>x</mi> </msub> <mrow> <msub> <mi>MVX</mi> <mrow> <mi>m</mi> <mi>e</mi> <mi>a</mi> <mi>n</mi> </mrow> </msub> </mrow> </mfrac> <mi>M</mi> <mi>V</mi> <msub> <mi>X</mi> <mrow> <mi>m</mi> <mi>e</mi> <mi>a</mi> <mi>n</mi> </mrow> </msub> <mo>&gt;</mo> <mn>0</mn> </mtd> </mtr> </mtable> </mfenced> <mo>;</mo> </mrow>
    <mrow> <msub> <mi>L</mi> <mi>y</mi> </msub> <mo>=</mo> <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mfrac> <mrow> <mi>H</mi> <mo>-</mo> <msub> <mi>C</mi> <mi>y</mi> </msub> </mrow> <mrow> <mo>-</mo> <msub> <mi>MVY</mi> <mrow> <mi>m</mi> <mi>e</mi> <mi>a</mi> <mi>n</mi> </mrow> </msub> </mrow> </mfrac> <mi>M</mi> <mi>V</mi> <msub> <mi>Y</mi> <mrow> <mi>m</mi> <mi>e</mi> <mi>a</mi> <mi>n</mi> </mrow> </msub> <mo>&lt;</mo> <mn>0</mn> </mtd> </mtr> <mtr> <mtd> <mfrac> <msub> <mi>C</mi> <mi>y</mi> </msub> <mrow> <msub> <mi>MVY</mi> <mrow> <mi>m</mi> <mi>e</mi> <mi>a</mi> <mi>n</mi> </mrow> </msub> </mrow> </mfrac> <mi>M</mi> <mi>V</mi> <msub> <mi>Y</mi> <mrow> <mi>m</mi> <mi>e</mi> <mi>a</mi> <mi>n</mi> </mrow> </msub> <mo>&gt;</mo> <mn>0</mn> </mtd> </mtr> </mtable> </mfenced> <mo>;</mo> </mrow>
    Wherein, Lx、LyThe existence time that respectively horizontal, vertical direction is estimated, W, H are respectively the width of video sequence, height, Cx、Cy Respectively current CTU horizontal stroke, ordinate, MVXmean、MVYmeanFor MVmeanX, y-component；Detect and include dynamic duplicate contents CTU after, in estimating in the range of existence time for dynamic duplicate contents, no longer detect new dynamic duplicate contents；
    It is spaced duplicate contents detection：When current reference piece is referred to by some CTU, then corresponding CTU is fast on current reference piece Total cost of fast estimation will be recorded；Meanwhile the position that current reference piece matches in fast motion estimation is also recorded； If current reference piece is repeatedly referred to, current reference piece needs the average and match bit for retaining the total cost of fast motion estimation The union put；When being encoded to M frames, for each CTU on M frames, reference plate retrieval is carried out in reference plate buffering area, such as The total cost of fast motion estimation that fruit obtains is less than certain times of the total cost average of fast motion estimation of current reference piece record Number, and the position matched includes the certain multiple of the matched position union of current reference piece record, then judges that corresponding CTU's is interior Hold and the reference plate that retrieves in content exist and repeat；The l if reference plate retrieved has been cited before M frames are compiled It is secondary, infer in follow-up cataloged procedure, similar duplicate contents there is also LCIt is secondary, LCCalculation be shown below：
    <mrow> <msub> <mi>L</mi> <mi>C</mi> </msub> <mo>=</mo> <mfrac> <mi>l</mi> <mi>M</mi> </mfrac> <mo>&amp;times;</mo> <mrow> <mo>(</mo> <mi>P</mi> <mo>-</mo> <mi>M</mi> <mo>)</mo> </mrow> <mo>;</mo> </mrow>
    Wherein, P is the length of video sequence.
  4. 4. a kind of video interprediction encoding method based on reference plate according to claim 1, it is characterised in that described Distribution situation generation based on high quality encoding block includes the reference plate of duplicate contents；Reference plate comprising duplicate contents directly adds Enter reference plate buffering area, then according to reference plate priority ratio be relatively managed including：
    Distribution situation generation based on high quality coding CTU includes the reference plate of duplicate contents, and its step is as follows：Encode current After frame, according to specified reference plate size, construction sliding window from top to bottom, from left to right scan current reference frame, find out window The intraoral CTU quantity for carrying out high quality coding reaches the position of maximum；If the maximum exceedes the threshold value specified, window is used Intraoral pixel value generation reference plate；The reference plate of generation is the reference plate for including duplicate contents, wherein also including following letter Breath：All CTU average qp value in pixel value, window；After the reference plate generation of duplicate contents, by the high quality of its present position Code identification is changed to regular quality coding, continues the scanning process of the above, and the ginseng of duplicate contents can be generated until can not select Untill examining piece；
    The reference plate comprising duplicate contents of generation is directly added into reference plate buffer area；Mode is as follows：If include duplicate contents Reference plate in, have the reference plate that one or more CTU with reference to, then with comprising duplicate contents reference plate replace ginseng Examine in piece buffering area that an existing reference plate, the reference plate being specifically replaced are currently to include the reference of duplicate contents All CTU choose that most reference plate of reference number of times in piece；Otherwise, reference plate is replaced with the reference plate comprising duplicate contents The minimum reference plate of priority in buffering area；
    Afterwards, the reference plate comprising duplicate contents of reference plate buffering area is relatively managed according to reference plate priority ratio.
  5. 5. a kind of video interprediction encoding method based on reference plate according to claim 1, it is characterised in that described Roughing goes out the reference plate of multiple candidates from reference plate buffering area by the way of Histogram Matching, then using fast motion estimation Mode the step of choosing a reference plate include：
    Histogram Matching：The histogram between all reference plates in current CTU to be encoded and reference plate buffering area is calculated first Distance, then result of the less reference plate of multiple Histogram distances as Histogram Matching is selected using predetermined way；
    The calculation of Histogram distance is as follows：1) its histogram is calculated to current CTU to be encoded：By current CTU's to be encoded Luminance component is divided into the block of 4 32 × 32, and the histogram of one H dimension of every piece of statistics is each to count to two chromatic components of U and V The histogram of one H dimension；Six H are tieed up into histogram and are spliced to form 6H dimension histograms；2) to each reference in reference plate buffering area The block of 64 × 64 sizes in piece, with level, several pixels are once sampled with vertical direction interval, to the block meter of sampling Calculate histogram；3) current CTU to be encoded histogram and the histogram of all sampling blocks of reference plate are contrasted one by one, calculated The distance between histogram, and wherein minimum distance is selected as current CTU to be encoded and the Histogram distance of reference plate；
    Fast motion estimation：In the multiple reference plates obtained from Histogram Matching, using the method for fast motion estimation, one is selected The individual reference plate most possibly utilized；Current CTU to be encoded carries out the calculation of fast motion estimation such as with each reference plate Under：Current CTU is divided into the fritter that fixed size is R × R, each R × R block search cost in current reference piece is minimum Block, wherein cost is made up of the encoder bit rate of absolute difference, sad value and motion vector；The mode of search is first by every 5 point Raster scanning determines the starting point searched in reference plate, is then carried out up and down in certain pixel coverage around the starting point 8 brilliant search；The cost of all R × R block is cumulative as current entirety of the CTU to be encoded on current reference piece Cost, and then filter out corresponding reference plate during overall Least-cost.
  6. A kind of 6. video interprediction encoding method based on reference plate according to claim 1 or 5, it is characterised in that Use predetermined mode to handle for new reference frame the reference plate chosen, and by its with original reference frame in Video coding simultaneously Row refer to for current block to be encoded to be included：
    Original reference frame in new reference frame and Video coding is arranged side by side, used when being encoded for current block to be encoded；Wherein, currently For block to be encoded when carrying out inter prediction encoding on new reference frame, the starting point searched for during estimation is set to Histogram Matching institute The optimum position found；Motion-vector prediction MVP choosing method is as follows：Check around current prediction unit PU on five positions Block whether employ reference plate, and if only if uses reference plate, and its reference plate with the reference plate used in current PU from same During one reference frame, then the motion vector MV on relevant position is as candidate MVP；Two MVP are therefrom chosen in order adds time MVP lists are selected, if the available MVP obtained on locus less than two, uses from the position where current block to be encoded successively The MV for pointing to reference plate center and (0,0) filling are put, untill filling up two.
  7. 7. a kind of video interprediction encoding method based on reference plate according to claim 1, it is characterised in that described The flag bit for the reference plate chosen whether will be with reference to for indicating the encoding block in current CTU to be encoded, and if with reference to When, the reference plate chosen is incorporated into code stream in the index of reference plate buffering area to be included：
    Whether with reference to the flag bit of reference plate chosen be one 0 or 1 flag bit, flag bit is using context-adaptive Whether binary arithmetic coding, the context model used have three, referred to by the CTU on current the CTU left sides and top to be encoded Reference plate determines, if two CTU are predicted without using reference plate, with first context model；If one of them is adopted With reference plate prediction, then with second context model；If both having used reference plate, with the 3rd context model；
    If current CTU to be encoded with reference to the reference plate chosen, it is fixed that the reference plate chosen uses in the index of buffering area Long codes, the reference plate number of reference plate buffering area when the reference plate that code length depends on choosing is referenced；
    Afterwards, by corresponding flag bit, code stream is incorporated into the index of buffering area with the reference plate chosen.
  8. A kind of 8. video interprediction encoding method based on reference plate according to claim 7, it is characterised in that the party Method also includes：Whether current CTU to be encoded has carried out high quality coding, and if carrying out high quality coding, high quality coding When QP relative to normal QP adjusted value, it is necessary to be incorporated into code stream；It is specific as follows：
    Whether carry out the flag bit of high quality coding be one 0 or 1 flag bit, flag bit two entered using context-adaptive Arithmetic coding processed, the context model used have three, high-quality by whether the CTU on current the CTU left sides and top to be encoded is carried out Amount is encoded to determine, if two CTU are no to carry out high quality coding, with first context model；If one of them enters Row high quality encodes, then with second context model；If high quality coding is both carried out, with the 3rd context mould Type；
    If carrying out high quality coding, QP when high quality encodes uses block code relative to normal QP adjusted value.
CN201610979281.9A 2016-11-08 2016-11-08 Video interprediction encoding method based on reference plate Active CN106507106B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610979281.9A CN106507106B (en) 2016-11-08 2016-11-08 Video interprediction encoding method based on reference plate

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610979281.9A CN106507106B (en) 2016-11-08 2016-11-08 Video interprediction encoding method based on reference plate

Publications (2)

Publication Number Publication Date
CN106507106A CN106507106A (en) 2017-03-15
CN106507106B true CN106507106B (en) 2018-03-06

Family

ID=58323882

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610979281.9A Active CN106507106B (en) 2016-11-08 2016-11-08 Video interprediction encoding method based on reference plate

Country Status (1)

Country Link
CN (1) CN106507106B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
RU2767993C1 (en) * 2018-05-16 2022-03-22 Хуавей Текнолоджиз Ко., Лтд. Video encoding device and method

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107295334B (en) * 2017-08-15 2019-12-03 电子科技大学 Adaptive reference picture chooses method
CN109640089B (en) * 2018-11-02 2023-03-24 西安万像电子科技有限公司 Image coding and decoding method and device
EP3886438A4 (en) * 2018-12-24 2022-01-19 Huawei Technologies Co., Ltd. Flag bit context modeling method and device
CN111464810A (en) * 2020-04-09 2020-07-28 上海眼控科技股份有限公司 Video prediction method, video prediction device, computer equipment and computer-readable storage medium
CN111901597B (en) * 2020-08-05 2022-03-25 杭州当虹科技股份有限公司 CU (CU) level QP (quantization parameter) allocation algorithm based on video complexity
CN114882390B (en) * 2022-03-15 2024-05-28 北京工业大学 Video frame type decision method based on CTU histogram in VVC coding standard
CN117615129B (en) * 2024-01-23 2024-04-26 腾讯科技（深圳）有限公司 Inter-frame prediction method, inter-frame prediction device, computer equipment and storage medium

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101461242A (en) * 2006-03-30 2009-06-17 Lg电子株式会社 A method and apparatus for decoding/encoding a video signal
CN102067608A (en) * 2008-06-25 2011-05-18 高通股份有限公司 Fragmented reference in temporal compression for video coding
CN102461171A (en) * 2009-05-01 2012-05-16 汤姆森特许公司 Reference picture lists for 3dv
CN104768017A (en) * 2014-01-03 2015-07-08 联发科技股份有限公司 Video coding method
CN104811729A (en) * 2015-04-23 2015-07-29 湖南大目信息科技有限公司 Multi-reference-frame encoding method for videos

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TWI530161B (en) * 2011-06-07 2016-04-11 Sony Corp Image processing apparatus and method
KR20140085456A (en) * 2011-09-29 2014-07-07 텔레폰악티에볼라겟엘엠에릭슨(펍) reference picture list handling

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101461242A (en) * 2006-03-30 2009-06-17 Lg电子株式会社 A method and apparatus for decoding/encoding a video signal
CN102067608A (en) * 2008-06-25 2011-05-18 高通股份有限公司 Fragmented reference in temporal compression for video coding
CN102461171A (en) * 2009-05-01 2012-05-16 汤姆森特许公司 Reference picture lists for 3dv
CN104768017A (en) * 2014-01-03 2015-07-08 联发科技股份有限公司 Video coding method
CN104811729A (en) * 2015-04-23 2015-07-29 湖南大目信息科技有限公司 Multi-reference-frame encoding method for videos

Non-Patent Citations (6)

* Cited by examiner, † Cited by third party
Title
A Unified framework of hash-based matching for screen content coding;bli li等;《IEEE》;20140817;全文 *
spatially scalable video coding for hevc;zhongbo shi等;《IEEE》;20121011;全文 *
steganography in modern smartphones and mitigation techniques;wojciech等;《IEEE》;20150325;全文 *
weighted rate-distortion optimization for screen content coding;WEIXIAO等;《IEEE》;20160123;全文 *
基于模型的编码;吴峰等;《计算机学报》;19991231;全文 *
渐进、精细的可伸缩性视频编码;吴峰等;《计算机学报》;20001231;全文 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
RU2767993C1 (en) * 2018-05-16 2022-03-22 Хуавей Текнолоджиз Ко., Лтд. Video encoding device and method

Also Published As

Publication number Publication date
CN106507106A (en) 2017-03-15

Similar Documents

Publication Publication Date Title
CN106507106B (en) Video interprediction encoding method based on reference plate
US11889072B2 (en) Video encoding and decoding
JP6073405B2 (en) Video decoding method and apparatus
CN103248893B (en) From H.264/AVC standard to code-transferring method and transcoder thereof the fast frame of HEVC standard
CN103220528B (en) Method and apparatus by using large-scale converter unit coding and decoding image
CN104796694B (en) Optimization intraframe video coding method based on video texture information
CN103765892B (en) The method and apparatus decoded using infra-frame prediction to the method and apparatus of Video coding and to video
US7672377B2 (en) Method and system for video encoding and transcoding
CN110521205A (en) Sub- predicting unit temporal motion vector prediction for coding and decoding video
CN102474609B (en) Method and apparatus for encoding images and method and apparatus for decoding images
EP1022667A2 (en) Methods of feature extraction of video sequences
CN106131546B (en) A method of determining that HEVC merges and skip coding mode in advance
CN104980761A (en) Method and device for coding and decoding motion vector
CN104980739A (en) Method And Apparatus For Video Encoding Using Deblocking Filtering, And Method And Apparatus For Video Decoding Using The Same
CN103563382A (en) Method and apparatus for encoding images and method and apparatus for decoding images
CN109104609A (en) A kind of lens boundary detection method merging HEVC compression domain and pixel domain
CN107277506B (en) Motion vector accuracy selection method and device based on adaptive motion vector precision
CN101453649A (en) Key frame extracting method for compression domain video stream
CN107079165A (en) Use the method for video coding and device of prediction residual
CN103596004A (en) Intra-frame prediction method and device based on mathematical statistics and classification training in HEVC
CN104604226A (en) Method and apparatus for coding video having temporal scalability, and method and apparatus for decoding video having temporal scalability
CN104811729B (en) A kind of video multi-reference frame coding method
CN107318016A (en) A kind of HEVC inter-frame forecast mode method for rapidly judging based on zero piece of distribution
CN103533355A (en) Quick coding method for HEVC (high efficiency video coding)
CN103384324A (en) Quick sub pixel motion estimation method for AVS-M video coding

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant