CN105611299B - A kind of method for estimating based on HEVC - Google Patents
A kind of method for estimating based on HEVC Download PDFInfo
- Publication number
- CN105611299B CN105611299B CN201510994165.XA CN201510994165A CN105611299B CN 105611299 B CN105611299 B CN 105611299B CN 201510994165 A CN201510994165 A CN 201510994165A CN 105611299 B CN105611299 B CN 105611299B
- Authority
- CN
- China
- Prior art keywords
- carried out
- pixel
- fractional
- texture
- steps
- 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
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
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/503—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction
- H04N19/51—Motion estimation or motion compensation
- H04N19/56—Motion estimation with initialisation of the vector search, e.g. estimating a good candidate to initiate a search
-
- 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
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/503—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction
- H04N19/51—Motion estimation or motion compensation
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
A kind of method for estimating based on HEVC, including step are disclosed：(1) integer pixel estimation is carried out to current PU；(2) judge whether current PU fractional-pel motion estimation is invalid, is, integer pixel MV is set to optimal MV and arrives step (9), it is no to then follow the steps (3)；(3) judge whether current PU includes textural characteristics, be to then follow the steps (4), otherwise use FSPS algorithm and arrive step (9)；(4) (7) are thened follow the steps if it is horizontal texture, (5) are thened follow the steps if it is vertical texture, integer pixel MV is then set to optimal MV if it is gridiron pattern texture and jumps to step (9), otherwise use FSPS algorithm and jumps to step (9)；(5) vertical direction interpolation is carried out；(6) vertical direction search is carried out, is arrived step (9)；(7) horizontal direction interpolation is carried out；(8) horizontal direction search is carried out；(9) terminate.
Description
Technical field
The invention belongs to the technical fields of HEVC Video coding, more particularly to a kind of estimation side based on HEVC
Method.
Background technique
HEVC (High Efficiency Video Coding, efficient video coding) is newest video encoding standard.
In HEVC coding standard, estimation can effectively remove the time redundancy of video sequence consecutive frame, largely determine
The coding rate of video encoder, compression ratio and decoded video quality.Therefore, HEVC motion estimation module increases a variety of volumes
Code technology, the brightness of 1/2 and 1/4 pixel is generated for example, by using one-dimensional 7 tap of fixation or filter of 8 taps based on DCT respectively
Value.The performance of HEVC has surmounted previous all video encoders, and under the premise of identical coding quality, HEVC is relative to H.264
50% code rate can be saved.The high-performance of HEVC is to rise to cost with computation complexity.
In HEVC encoder, estimation includes whole pixel and fraction pixel two parts.Fractional-pel motion estimation be
Obtain whole pixel motion vector (MV:Motion vector) on the basis of carry out interpolation arithmetic, search obtains fraction pixel precision
The process of MV.Fractional-pel motion estimates the performance that encoder can be greatlyd improve on compressed video quality and compression ratio.
But due to additional operation, such as interpolation and fractional pixel search, fractional-pel motion estimation substantially increases entire movement
The calculation amount of estimation module.
The fractional-pel motion algorithm for estimating that HEVC is used is fraction pixel full-search algorithm (FSPS：Full Sub-
pixel Search).As shown in Figure 1, I is the integer pixel positions obtained by integer pixel estimation, h1, h2, h3 ...,
H8 is 8 half-pixel positions around I.FSPS algorithm calculates the absolute error and (SAD of this 8 positions:sum of
Absolute difference), and therefrom select the smallest position SAD.The SAD of h2 is minimum in Fig. 1.Then, it calculates optimal
8 a quarter location of pixels q1, q2, q3 around half-pixel position h2 ... the SAD of q8, and therefrom select optimal location.q8
SAD it is minimum, so q8 is final optimal location.
Fig. 2 is HEVC brightness layer fractional pixel interpolation schematic diagram.ai,j、bi,j、ci,j、di,j、hi,jAnd ni,jOnly by integer picture
Plain interpolation can obtain.And to obtain ei,j、fi,j、gi,j、ii,j、ji,j、ki,j、pi,j、qi,jAnd ri,j, first have to calculate ai,j,
bi,jAnd ci,j, interpolation then is carried out to them in the vertical direction, has thus obtained all fraction pixels.
Estimation is the key technology of various video encoding standards.It is proposed with whole pixel algorithm with continuously improve,
Optimal solution can be found in the case where only detecting 5 Searching points.In contrast, the search finding calculation amount of fraction pixel is just
Seem and be difficult to receive, therefore develop the fast motion estimation algorithm of fraction precision, there is urgent need.
Summary of the invention
Technology of the invention solves the problems, such as：Overcome the deficiencies of the prior art and provide a kind of estimation based on HEVC
Method can reduce the complexity of estimation under the premise of not reducing coding quality, thus computation amount, meter
Speed is calculated to greatly improve.
The technical solution of the invention is as follows：This method for estimating based on HEVC, includes the following steps：
(1) integer pixel estimation is carried out to current prediction unit PU；
(2) judge whether current PU fractional-pel motion estimation is invalid, is that integer pixel MV is set to optimal MV simultaneously
It jumps to step (9), it is no to then follow the steps (3)；
(3) judge whether current PU includes textural characteristics, be to then follow the steps (4), otherwise use FSPS algorithm and jump
To step (9)；
(4) detection of texture classification is carried out, (7) is thened follow the steps if it is horizontal texture, is then executed if it is vertical texture
Integer pixel MV is then set to optimal MV if it is gridiron pattern texture and jumped to step (9) by step (5), if current PU
There is no the above textural characteristics, then use FSPS algorithm and jumps to step (9)；
(5) vertical direction interpolation is carried out；
(6) vertical direction search is carried out, step (9) are jumped to；
(7) horizontal direction interpolation is carried out；
(8) horizontal direction search is carried out；
(9) terminate.
By judging integer pixel estimation effect, invalid score picture is skipped in region good enough to one aspect of the present invention
On the other hand plain motion search determines different fractional pixel search methods according to texture analysis, therefore is not reducing coding matter
Reduce the complexity of estimation under the premise of amount, thus computation amount, calculating speed is greatly improved.
Detailed description of the invention
Fig. 1 shows the fractional-pel motion algorithm for estimating of HEVC in the prior art.
Fig. 2 is HEVC brightness layer fraction pixel schematic diagram.
Fig. 3 a shows the fractional pixel search algorithm based on vertical texture, and Fig. 3 b shows the score based on horizontal texture
Pixel search algorithm.
Fig. 4 is the flow chart of the method for estimating according to the present invention based on HEVC.
Specific embodiment
As shown in figure 4, this method for estimating based on HEVC, includes the following steps：
(1) integer pixel estimation is carried out to current prediction unit PU；
(2) judge whether current PU fractional-pel motion estimation is invalid, is that integer pixel MV is set to optimal MV simultaneously
It jumps to step (9), it is no to then follow the steps (3)；
(3) judge whether current PU includes textural characteristics, be to then follow the steps (4), otherwise (fraction pixel is complete using FSPS
Searching algorithm, Full Sub-pixel Search) algorithm and jump to step (9)；
(4) detection of texture classification is carried out, (7) is thened follow the steps if it is horizontal texture, is then executed if it is vertical texture
Integer pixel MV is then set to optimal MV if it is gridiron pattern texture and jumped to step (9) by step (5), if current PU
There is no the above textural characteristics, then use FSPS algorithm and jumps to step (9)；
(5) vertical direction interpolation is carried out；
(6) vertical direction search is carried out, step (9) are jumped to；
(7) horizontal direction interpolation is carried out；
(8) vertical direction search is carried out；
(9) terminate.
By judging integer pixel motion search effect, invalid score picture is skipped in region good enough to one aspect of the present invention
On the other hand plain motion search determines different fractional pixel search methods according to texture analysis, therefore is not reducing coding matter
Reduce the complexity of estimation under the premise of amount, thus computation amount, calculating speed is greatly improved.
Preferably, it is determined that whether current PU fractional-pel motion is estimated to be invalid with formula (1) in the step (2)
's：
Wherein, MVIntIt is integer pixel MV, T is MAE (MVInt) threshold value, MAE (MVInt) definition be formula (2)
Wherein, SAD (MVInt) be the corresponding absolute error of integer pixel MV and, width is the width of current PU, height
It is the height of current PU, T is obtained according to formula (3)：
T=-0.014 × qp2+1.095×qp-14.07 (3)
Wherein, qp is quantization parameter.
Preferably, the fractional pixel position that vertical direction interpolation obtains only is searched through in the step (6).
Preferably, the fractional pixel position that horizontal direction interpolation obtains only is searched through in the step (8).
The present invention will be described in more detail below.
1. selectively skipping invalid fractional pixel search
Compressed video quality and compression ratio can be improved in fractional-pel motion estimation, but fraction pixel MV search needs are huge
Big calculation amount.If least absolute error and (MSAD that score MV search obtains:minimum sum of absolute
Difference) it is equal to the MSAD that whole pixel MV search obtains, then whole pixel MV is selected to search for as final MV, fraction pixel MV
It is considered as invalid., whereas if the MSAD that score MV search obtains is less than the MSAD, fraction pixel MV that whole pixel MV search obtains
It is effective for searching for, and selects fraction pixel MV as final MV.30 frames before 6 standard video sequences are tested, it will be final
MV is that ratio shared by whole pixel MV is shown in Table 1.
Ratio (30 frames, QP=27) shared by the whole pixel MV of table 1
According to the experimental results, the MV for averagely having more than 60% is located at whole location of pixels.In fact, certain for video frame
A little regions, fractional pixel search are not obvious the raising of coding efficiency, are invalid search.If these integer pictures can be predicted
The plain sufficiently exact region unit of search result, skips invalid fractional pixel search, so that it may reduce unnecessary calculating.
The sufficiently exact predicting unit of integer-pixel search result of video frame is judged before fractional-pel motion estimation
(PU：Prediction unit) it is to skip the key of invalid fractional-pel motion search.
It can determine whether to skip fractional-pel motion with the corresponding mean absolute error of integer pixel MV of current PU
Search.Skip_fraction_flag is whether to skip the mark of invalid fractional-pel motion estimation.Its definition is
Wherein MVIntIt is integer pixel MV, T is MAE (MVInt) threshold value.MAE(MVInt) definition be
Wherein SAD (MVInt) it is the corresponding absolute error of integer pixel MV and (SAD：sum of absolute
Difference), width is the width of current PU, and height is the height of current PU.The definition of threshold value T is proposed by experiment：
T=-0.014 × qp2+1.095×qp-14.07 (3)
Wherein qp is quantization parameter (QP, Quantization Parameter).
2. the fractional-pel motion based on grain direction is estimated
If current PU includes vertical texture, the interpolation for carrying out horizontal direction to the integer pixel of current PU be will lead to
Video blur.Similarly, for the PU comprising horizontal texture, the interpolation for carrying out vertical direction to its integer pixel also results in view
Frequency is fuzzy.So we can skip some by obtaining with texture opposite direction interpolation for there is the PU of horizontal or vertical direction texture
The fractional pixel position arrived.
Firstly, analyzing vertical texture.For vertical texture, it is only necessary to be searched through the score picture that vertical direction interpolation obtains
Plain position.In Fig. 2, only di,j、hi,jAnd ni,jIt is to be obtained by vertical direction interpolation.As shown in Fig. 3 (a), biggish pros
Shape I is the integer pixel positions obtained by integer pixel motion search, and 8 small squares are 8 half-pixel positions around I.
The SAD for only calculating two positions A, B, then finds an optimal position.Assuming that the SAD of A is minimum, then optimal half-pix
Two a quarter location of pixels of vertical direction are 1,2 around the A of position.So to calculate 1,2 two a quarter pixel
The SAD of position.Finally select the smallest position of sad value.
Secondly, to discuss the PU with horizontal texture.For horizontal texture, it is only necessary to be searched through horizontal direction interpolation and obtain
The fractional pixel position arrived.In Fig. 2, only ai,j、bi,jAnd ci,jIt is to be obtained by vertical direction interpolation.As shown in Fig. 3 (b),
Biggish square I is the integer pixel positions obtained by integer pixel motion search, and 8 small squares are 8 half around I
Location of pixels.The SAD for only calculating two positions C, D, then finds an optimal position.Assuming that the SAD of C is minimum, then most
Two a quarter location of pixels of vertical direction are 1,2 around excellent half-pixel position C.So to calculate 1,2 two four
The SAD of/mono- location of pixels.Finally select the smallest position of sad value.
Finally, can directly be skipped if current PU is chessboard trellis texture (existing horizontal texture has vertical texture) again
Fractional-pel motion estimation.Because in addition to a in Fig. 2i,j、bi,j、ci,j、di,j、hi,jAnd ni,jAll positions be all by level
What interpolation and vertical interpolation obtained jointly.Any one in both interpolation can all lead to video blur.
3. the major architectural of motion estimation algorithm
Based on analysis above, the major architectural of motion estimation algorithm is proposed, as shown in Figure 4.Specifically, current PU
Estimation process it is as described below.Firstly, carrying out integer pixel estimation to current PU.It is determined with formula (1) to working as
Whether preceding PU fractional-pel motion estimation is invalid.If fractional-pel motion estimation is invalid to current PU, skip point
Number pixel motion estimation, and integer pixel MV is set to optimal MV.If fractional-pel motion estimation is effectively, need to detect
The textural characteristics of current PU.If current PU includes chessboard trellis texture, fractional-pel motion estimation is skipped, and whole
Number pixel MV is set to optimal MV.For vertical texture, it should using vertical interpolation and vertical search.For horizontal texture, answer
This uses Horizontal interpolation and horizon scan.If current PU does not have above-mentioned textural characteristics, the FSPS algorithm of HEVC is used.
Above scheme is had been applied in HEVC encoder, and achieves apparent effect.
The experimental results showed that fraction pixel fast motion estimation algorithm is than fraction pixel full-search algorithm (FSPS) in peak value
In the case that signal-to-noise ratio (PSNR) has small reduction (0.01dB), bitrate averagely increases 0.02%, and total scramble time is flat
Reduce 24.56%, entire motion estimation time averagely saves 40.86%.
Table 2
The above is only presently preferred embodiments of the present invention, is not intended to limit the present invention in any form, it is all according to
According to technical spirit any simple modification, equivalent change and modification to the above embodiments of the invention, still belong to the present invention
The protection scope of technical solution.
Claims (3)
1. a kind of method for estimating based on HEVC, it is characterised in that：Include the following steps：
(1) integer pixel estimation is carried out to current prediction unit PU；
(2) judge whether current PU fractional-pel motion estimation is invalid, is that integer pixel MV is set to optimal MV and is jumped
It is no to then follow the steps (3) to step (9)；
(3) judge whether current PU includes textural characteristics, be to then follow the steps (4), otherwise use FSPS algorithm and jump to step
Suddenly (9)；
(4) detection of texture classification is carried out, (7) is thened follow the steps if it is horizontal texture, is thened follow the steps if it is vertical texture
(5), integer pixel MV is then set to optimal MV if it is gridiron pattern texture and jumped to step (9), if current PU does not have
The above textural characteristics then use fraction pixel full-search algorithm FSPS and jump to step (9)；
(5) vertical direction interpolation is carried out；
(6) vertical direction search is carried out, step (9) are jumped to；
(7) horizontal direction interpolation is carried out；
(8) horizontal direction search is carried out；
(9) terminate；
It is determined that whether current PU fractional-pel motion is estimated to be invalid with formula (1) in the step (2)：
Wherein, MVIntIt is integer pixel MV, T is MAE (MVInt) threshold value, MAE (MVInt) definition be formula (2)
Wherein, SAD (MVInt) be the corresponding absolute error of integer pixel MV and, width is the width of current PU, and height is to work as
The height of preceding PU, T are obtained according to formula (3)：
T=-0.014 × qp2+1.095×qp-14.07 (3)
Wherein, qp is quantization parameter.
2. the method for estimating according to claim 1 based on HEVC, it is characterised in that：In the step (5) only
It carries out vertical direction interpolation and obtains fractional pixel position, be only searched through what vertical direction interpolation obtained in the step (6)
Fractional pixel position.
3. the method for estimating according to claim 2 based on HEVC, it is characterised in that：In the step (7) only
It carries out horizontal direction interpolation and obtains fractional pixel position, be only searched through what horizontal direction interpolation obtained in the step (8)
Fractional pixel position.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510994165.XA CN105611299B (en) | 2015-12-25 | 2015-12-25 | A kind of method for estimating based on HEVC |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510994165.XA CN105611299B (en) | 2015-12-25 | 2015-12-25 | A kind of method for estimating based on HEVC |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105611299A CN105611299A (en) | 2016-05-25 |
| CN105611299B true CN105611299B (en) | 2018-11-23 |
Family
ID=55990796
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510994165.XA Active CN105611299B (en) | 2015-12-25 | 2015-12-25 | A kind of method for estimating based on HEVC |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105611299B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020003258A1 (en) | 2018-06-29 | 2020-01-02 | Beijing Bytedance Network Technology Co., Ltd. | Virtual merge candidates |
| WO2022061613A1 (en) * | 2020-09-23 | 2022-03-31 | 深圳市大疆创新科技有限公司 | Video coding apparatus and method, and computer storage medium and mobile platform |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1688165A (en) * | 2005-06-09 | 2005-10-26 | 上海交通大学 | Fast motion assessment method based on object texture |
| US6996180B2 (en) * | 2001-09-05 | 2006-02-07 | Intel Corporation | Fast half-pixel motion estimation using steepest descent |
| CN101699865A (en) * | 2009-10-28 | 2010-04-28 | 山东大学 | Method for fast searching mass movement self-adapting sub pixel |
| CN101795409A (en) * | 2010-03-03 | 2010-08-04 | 北京航空航天大学 | Content adaptive fractional pixel motion estimation method |
| CN104811728A (en) * | 2015-04-23 | 2015-07-29 | 湖南大目信息科技有限公司 | Video content self-adaptation motion search method |
-
2015
- 2015-12-25 CN CN201510994165.XA patent/CN105611299B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6996180B2 (en) * | 2001-09-05 | 2006-02-07 | Intel Corporation | Fast half-pixel motion estimation using steepest descent |
| CN1688165A (en) * | 2005-06-09 | 2005-10-26 | 上海交通大学 | Fast motion assessment method based on object texture |
| CN101699865A (en) * | 2009-10-28 | 2010-04-28 | 山东大学 | Method for fast searching mass movement self-adapting sub pixel |
| CN101795409A (en) * | 2010-03-03 | 2010-08-04 | 北京航空航天大学 | Content adaptive fractional pixel motion estimation method |
| CN104811728A (en) * | 2015-04-23 | 2015-07-29 | 湖南大目信息科技有限公司 | Video content self-adaptation motion search method |
Non-Patent Citations (1)
| Title |
|---|
| 《基于H.264的内容自适应分数像素运动估计算法》;祝世平等;《光电子·激光》;20100930;第21卷(第9期);第1369-1373页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105611299A (en) | 2016-05-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111385569B (en) | Coding and decoding method and equipment thereof | |
| US9706202B2 (en) | Image encoding apparatus, image encoding method, image decoding apparatus, and image decoding method | |
| CN101385356B (en) | Process for coding images using intra prediction mode | |
| EP2164266B1 (en) | Moving picture scalable encoding and decoding method using weighted prediction, their devices, their programs, and recording media storing the programs | |
| CN102100068B (en) | Speculative start point selection for motion estimation iterative search | |
| CN107623850B (en) | A kind of quick screen contents encoding method based on temporal correlation | |
| CN108134939B (en) | Motion estimation method and device | |
| TWI502966B (en) | Video encoding method and apparatus, video decoding method and apparatus, and programs thereof | |
| JP2008252176A (en) | Motion picture encoder and encoding method | |
| CN107360419B (en) | A kind of movement forward sight video interprediction encoding method based on perspective model | |
| CN102100059B (en) | Simple next search position selection for motion estimation iterative search | |
| CN103168470A (en) | Video encoding method, video decoding method, video encoding device, video decoding device, and programs for same | |
| TWI504241B (en) | Video encoding method and apparatus, video decoding method and apparatus, and programs product thereof | |
| CN103327327B (en) | For the inter prediction encoding unit selection method of high-performance video coding HEVC | |
| CN1604653A (en) | Differential video coding method | |
| CN110519591A (en) | A kind of prediction mode fast selecting method based on intraframe coding in multipurpose coding | |
| CN105611299B (en) | A kind of method for estimating based on HEVC | |
| CN109120940B (en) | Adaptive factor video scaling motion estimation method | |
| CN100542299C (en) | The concealing method of video image error | |
| CN105791865A (en) | Intra-frame prediction and deblocking filtering method | |
| CN113727118B (en) | Decoding method, encoding method, device, equipment and machine readable storage medium | |
| Wang et al. | Adaptive motion vector resolution prediction in block-based video coding | |
| KR20080022843A (en) | Method for adjusting search area using motion vector | |
| JP4222274B2 (en) | Encoding mode selection device and encoding mode selection program | |
| Wong et al. | Sub-optimal quarter-pixel inter-prediction algorithm (SQIA) |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |