CN103905818B - Method for rapidly determining inter-frame prediction mode in HEVC standard based on Hough conversion - Google Patents
Method for rapidly determining inter-frame prediction mode in HEVC standard based on Hough conversion Download PDFInfo
- Publication number
- CN103905818B CN103905818B CN201410161592.5A CN201410161592A CN103905818B CN 103905818 B CN103905818 B CN 103905818B CN 201410161592 A CN201410161592 A CN 201410161592A CN 103905818 B CN103905818 B CN 103905818B
- Authority
- CN
- China
- Prior art keywords
- edge
- current
- value
- point
- rmd
- 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
- 238000000034 method Methods 0.000 title claims abstract description 122
- 238000006243 chemical reaction Methods 0.000 title abstract description 6
- 230000008569 process Effects 0.000 claims abstract description 42
- 238000001514 detection method Methods 0.000 claims abstract description 18
- 238000007906 compression Methods 0.000 claims abstract description 3
- 230000006835 compression Effects 0.000 claims abstract description 3
- 238000004364 calculation method Methods 0.000 claims description 10
- 238000000605 extraction Methods 0.000 claims description 9
- 230000000877 morphologic effect Effects 0.000 claims description 7
- 235000021061 dietary behavior Nutrition 0.000 claims description 6
- 235000006694 eating habits Nutrition 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 6
- 238000007670 refining Methods 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 3
- 238000005260 corrosion Methods 0.000 claims description 2
- 230000007797 corrosion Effects 0.000 claims description 2
- 230000003628 erosive effect Effects 0.000 claims description 2
- 241001212149 Cathetus Species 0.000 claims 1
- 238000004458 analytical method Methods 0.000 abstract description 2
- 238000003708 edge detection Methods 0.000 abstract description 2
- 238000002474 experimental method Methods 0.000 abstract description 2
- 238000012360 testing method Methods 0.000 description 9
- 238000005516 engineering process Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000009977 dual effect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 239000011800 void material Substances 0.000 description 2
- 238000005530 etching Methods 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 238000012812 general test Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 238000010561 standard procedure Methods 0.000 description 1
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention discloses a method for rapidly determining an inter-frame prediction coding mode in the latest national video compression standard, namely, the HEVC standard. The method includes the steps of firstly, conducting edge detection and Hough conversion on a current PU; secondly, conducting statistic analysis on the tangent value of the direction angle of a detected straight line so as to extract the most possible edge direction of the current PU according to the heuristic type information, optimally selecting candidate prediction modes so as to determine the MPM by conducting detection through the code rate and RMD process, and determining the final coding mode by conducting detection through the code rate and RDO process. Experiments show that the running speed of the HEVC can be increased by more than 20% according to the method.
Description
Technical field
The present invention relates to accelerating it in up-to-date in the world at present compression standard HEVC in field of video encoding to run speed
The method of degree, more particularly, to a kind of optimal coding mode accelerating its Intra-coded blocks using technology such as Hough (Hough) conversion
Determination method.
Background technology
HEVC (High Efficiency Video Coding) is International video coding standard up-to-date at present.It is by
ISO/IEC mobile image expert group (MPEG) and ITU-T Video Coding Experts Group (VCEG) combine formulation.It is to continue H.264/
High-performance video coding standard after AVC.H.264, its code efficiency ratio improves more than 50%.HEVC still adopts pre-
Survey plus conversion hybrid encoding frame, the technology in this framework relatively before Technical comparing from the point of view of although coding framework substantially
Content and flow process do not have excessive change, but in order to improve video coding efficiency, add many new in HEVC standard yet
Technology.
In H.264 video encoding standard, macro block (Micro Block, MB) is the most basic unit of coding.But
In HEVC coding standard, due to the popularization of HD video, to the super clear video reaching 2K, 4K, if again using so fixing big
If little encoding block, it is difficult to adapt to the demand of HD video.Therefore, HEVC does not continue the concept using MB, but adopts
More flexi mode carries out block structure expression：Coding unit (CU coding unit), predicting unit (PU prediction
Unit), converter unit (TU transform unit).CU is most basic cutting unit, its maximum coding size is 64 ×
64, minimum code size is 8 × 8.PU is an elementary cell carrying and predicting process-related information, and PU is smaller in size than or waits
In CU size.One CU can comprise one or more various sizes of predicting unit PU, and a PU comprises some converter units
TU, TU are elementary cells based on conversion and quantization.HEVC carries out the division of coding unit using quad-tree structure.
HEVC in order to overcome H.264/AVC that middle predictive mode number is few, precision of prediction is not high, inaccurate the shortcomings of,
HEVC considerably increases the quantity of adoptable predictive mode.HEVC infra-frame prediction has 35 kinds of predictive modes, wherein pattern 0 table
Show and predicted using plane (Planar) mode, pattern 1 represents to be predicted using direct current (DC) mode, and pattern 2 to 34 represents using each
Plant angle to be predicted.
In the reference model HM4.0 of HEVC, the decision making process of frame mode is divided into three phases.First, to suitable difference
The set of patterns of size PU is roughly selected, calculate be possible to predictive mode rate distortion costs (Rate-Distortion Cost,
Rd-Cost), and by it according to cost descending order, one group of minimum predictive mode of Rd-Cost value is to roughly select subset,
This process is called pattern of roughly selecting and determines (RMD) process.Here, during calculated distortion, only consider the distortion of prediction residual, rather than
Distortion after encoding and decoding.Secondly, roughly select son with whether the intra prediction mode of upside reference block is included on the left of checking current block
Concentrating, if not existing, being added into subset.Finally, rate-distortion optimization, Least-cost are carried out to the predictive mode roughly selected in subset
Pattern be optimal prediction modes, this process is called code check and distortion factor optimization (RDO) process.Here, need during RDO
Calculate the distortion after the encoding and decoding of each candidate pattern, therefore its amount of calculation is very big.Much optional due to having in HEVC
Infra-frame prediction pattern, the determination process of its frame mode has very big computation complexity, very time-consuming.
In sum, with the optional predictive mode of frame in HEVC quantitative increase although improve precision of prediction and
Code efficiency, but also considerably increase the encoder complexity of HEVC simultaneously.Thus, frame in optimal prediction modes really fix time by
Substantially increase.
The present invention makes every effort to, on the basis of keeping precision of prediction, reduce intra prediction mode in HEVC and really fix time.By
Marginal information in the straight line in image block can bring the direction of the tendency of pixel in block to us, and the present invention will be opened using this
Hairdo information is come the intra prediction mode to simplify with further reduce candidate.
Content of the invention
In view of the drawbacks described above of prior art, the present invention proposes a kind of fast intra mode based on Hough transform
Determination new method.For realizing the purpose of the precision keeping as much as possible predicting while reducing amount of calculation, the present invention will
First rim detection is carried out to image, then adopt the edge that Hough transform search is straight line, next determine that present image is pre-
Survey the angle of the linear edge in unit PU, determine the most possible prediction direction of current PU using statistic histogram, and according to
This direction determines the predictive mode of most possible candidate, to determine (RMD) process RMD and code check and the distortion factor in pattern of roughly selecting
Reduce candidate pattern to be detected during optimizing (RDO), greatly reduce amount of calculation, and current PU is selected as much as possible simultaneously
The predictive mode selected.It is characterized in that, for each intraprediction unit PU in video image, methods described includes：
Step one, to current PU, judges whether its size is 4 × 4.In this way, then carry out the pattern sub-sampling that step-length is 1,
And leap to step 8.Otherwise, using Canny operator, row operation is entered to current PU.
Step 2, the result of the computing according to Canny operator, with morphologic method, edge is refined.
Step 3, selects suitable threshold value to carry out the extraction of marginal point and connection in PU.
Step 4, to current PU, using Hough transform detection of straight lines edge pixel point, i.e. retains and is on a certain straight line
Marginal point, remove and be in marginal point on curve.
Step 5, to each linear edge, determines the differential seat angle θ in itself and horizontal direction, and its tangent value tan θ.
Step 6, in the scope of the tangent value of the pre- measuring angle according to each intra prediction mode, and the current PU detecting
The tangent value of each linear edge, with statistic histogram determine linear edge pixel that current PU declines in the range of each it
With.
Step 7, according to this statistic histogram, determines several predictions of the linear edge pixel sum maximum in rectangular histogram
Pattern.
Step 8, according to the size of current PU, and the statistic histogram obtaining, determine Dietary behavior rougher process RMD
The frame mode of candidate.
Step 9, according to RMD process, and the size of current PU, determine the frame in of the candidate of Dietary behavior refining process RDO
Pattern.
Step 10, according to the result of RDO, determines the intra prediction mode of the optimization of current PU.
Further, in described step one, the block to 4 × 4 enters row mode sub-sampling.In HEVC reference model HM4.0
Totally 19 kinds of intra prediction mode being suitable for for 4 × 4PU, directional prediction pattern is respectively：7、14、6、13、1、12、5、11、
4、15、8、16、2、17、9、18、10.Block to 4 × 4 enters the pattern that row mode sub-sampling only retains half, that is, select mode 7,
6、1、5、4、8、2、9、10.
Canny operator edge detection step is as follows：Apply Gaussian filter smoothed image first, asked in 3 × 3 neighborhoods
The amplitude of limit difference mean value computation shade of gray and direction.
Further, in described step 2, carried out carefully using the morphologic method edge thick to the width detecting
Change, i.e. carry out morphologic erosion operation, the template elements of its corrosion are
Further, in described step 3, using threshold value ThCarry out marginal point extraction.Optimal threshold is chosen using iterative method
Value.It is as follows that iterative method realizes step：
(1) obtain minima Z in the current PU of imagemin, maximum Zmax, then Initial Hurdle T0=(Zmin+Zmax)/2.
(2) according to threshold values Tk(k is iterationses) divides the image into into marginal area and non-edge, calculates two parts
Meansigma methodss ZO、ZB, that is,
Here, hiAnd hjRepresent the value of the pixel of edge and non-edge respectively.Calculate ZO、ZBAfterwards, use formula
Tk+1=(ZO+ZB)/2
To calculate new threshold values Tk+1.
(3) if Tk+1=Tk, i.e. TkFor required threshold value, then algorithm leaves it at that, and otherwise goes to step (2).Iterative calculation is straight
To converging on certain stable threshold value, this threshold value is final result Th.
Here, detected from candidate marginal using dual threshold method and connect final edge.In the extraction of marginal point,
The threshold value being adopted is Th, the threshold value adopting during the connection of marginal point should compare ThLittle, with adjoining edge.At edge in the present invention
The threshold value adopting during the connection of point is Th/2.That is, after processing through the first and second steps, value is more than ThPixel be marginal point.So
Afterwards with these marginal points as seed, find the value being adjacent and be more than Th/ 2 pixel, adds as edge pixel point.Then, weight
This process multiple, in PU, all of pixel all scans and goes over.
The pre- measuring angle of each predictive mode of table 1. 8 × 8,16 × 16,32 × 32 size PU, boundary angles scope and
Tangent scope
The pre- measuring angle of each predictive mode of table 2. 64 × 64 size PU, boundary angles scope and tangent scope
| Predictive mode | Pre- measuring angle | Boundary angles scope | Tangent value scope |
| 1 | -90 | (- 45, -135) | (- ∞, -1) U (1 ,+∞) |
| 2 | 0 | [- 45,45] | [- 1,1] |
Further, in step 6, to the scope of statistics of the tangent value of linear edge angle as shown in Table 1 and Table 2.Example
As, as shown in table 1, for candidate pattern 7, the scope of statistics of its angle tangent value be (- 0.7417, -1.1033].I.e., initially
This in the range of linear pixel point value set to 0, C7=0, then the tangent value if there are the angle of linear edge fall in this scope,
Then C7=C7+ai, wherein aiIt is all pixels points on this straight line in current PU.All of candidate pattern is all carried out
This is processed, then obtained the statistic histogram of desired each pattern.Because pattern 0, pattern 3 are non-directional prediction patterns, therefore
Not by statistics in table 1, table 2, they are all entered into roughly selecting of RDO as candidate pattern.
The number of the predictive mode of candidate of RMD rougher process is entered in table 3.HM4.0 each PU size
| 4×4 | 19 |
| 8×8 | 35 |
| 16×16 | 35 |
| 32×32 | 35 |
| 64×64 | 4 |
The number of the predictive mode of candidate of RMD rougher process is entered in table 4. present invention each PU size
| 4×4 | 9 |
| 8×8 | 7 |
| 16×16 | 5 |
| 32×32 | 3 |
| 64×64 | 1 |
Further, the process of step 8 is as shown in table 4.For example, for the PU of 16 × 16 sizes, will select 5 with
The maximum candidate pattern of upper statistic histogram intermediate value.From the contrast of table 3 and table 4 it can be seen that present invention substantially reduces into
Enter the number of the pattern of candidate to RMD rougher process.As such, it is possible to greatly reduce amount of calculation, accelerate the fortune of HEVC standard
Scanning frequency degree.
The number of the predictive mode of candidate of RDO refining process is entered in table 5.HM4.0 each PU size
| 4×4 | 8 |
| 8×8 | 8 |
| 16×16 | 3 |
| 32×32 | 3 |
| 64×64 | 3 |
The number of the predictive mode of candidate of RDO refining process is entered in table 6. present invention each PU size
| 4×4 | 5 |
| 8×8 | 4 |
| 16×16 | 3 |
| 32×32 | 2 |
| 64×64 | 1 |
Further, the process of step 9 is as shown in table 6.For example, for the PU of 16 × 16 sizes, will select in RMD mistake
In journey, 3 optimum candidate pattern enter in the selection course of final RDO.From the contrast of table 5 and table 6 it can be seen that
Present invention substantially reduces the number of the pattern of the candidate entering into RDO refining process.As such, it is possible to greatly reduce calculating
Amount, accelerates the speed of service of HEVC standard.
In sum, the present invention proposes a kind of new side of the infra-frame prediction in up-to-date coding standard HEVC at present in the world
Method, to improve its speed of service.The innovative point of the method is：Carry out rim detection using Canny operator, and utilize Hough
Conversion carries out Straight edge inspection, according to these heuristic informations, to extract current prediction unit most probable edge trend.
Then, walk, according to these edges, the predictive mode always selecting and reducing the candidate needing detection.With this, to greatly reduce HEVC
Coding computation complexity, improve its speed of service.
Technique effect below with reference to design, concrete structure and generation to the present invention for the accompanying drawing is described further, with
It is fully understood from the purpose of the present invention, feature and effect.
Brief description
Fig. 1 is the schematic diagram of the intra prediction mode in HEVC standard；
Fig. 2 is the flow chart of new method proposed by the invention；
Fig. 3 is method proposed by the invention and original method (HM4.0) to video test sequence Party Scene (832
× 480p resolution) code check and distortion factor performance curve comparison diagram；
Fig. 4 is method proposed by the invention and original method (HM4.0) to video test sequence Slide Editing
The code check of (720p resolution) and distortion factor performance curve comparison diagram；
Fig. 5 is that (1080p divides to video test sequence Kimono1 for method proposed by the invention and original method (HM4.0)
Resolution) code check and distortion factor performance curve comparison diagram；
Fig. 6 is method proposed by the invention and original method (HM4.0) to video test sequence
The code check of SteamLocomotiveTrain (4k × 2k resolution) and distortion factor performance curve comparison diagram；
Specific embodiment
Below in conjunction with the accompanying drawings embodiments of the invention are elaborated：The present embodiment is with technical solution of the present invention premise
Under implemented, give detailed embodiment and specific operating process, but protection scope of the present invention be not limited to following
Embodiment.
It can be seen that the infra-frame prediction of original HEVC has a lot of candidate pattern from accompanying drawing 1, determine the mould of relatively optimization
Formula has larger computation complexity.
The present invention utilizes C Plus Plus on the basis of the program of the HM4.0 of HEVC, using OpenCV function library, to being carried
The method going out is realized and has been tested.
As shown in Figure 2, the method for the computation complexity of its infra-frame prediction of reduction of the present invention comprises the steps：
Step one, to current PU, judges whether its size is 4 × 4.In this way, then carry out the pattern sub-sampling that step-length is 1,
And leap to step 8.Otherwise, using Canny operator, row operation is entered to current PU.For the PU of 4 × 4 sizes, because its chi
Very little too little, the linear edge direction therefrom extracting and the dependency in the direction of predictive mode are not especially big, therefore directly jump
To step 8.Size is more than to the PU of 4 × 4 sizes, the linear edge direction therefrom extracting and the direction of predictive mode
Dependency is very big, therefore may be used to determine candidate pattern to be detected.OpenCV is accessed by function cvCanny
Canny operator carries out rim detection, and its function call prototype is void cvCanny (const CvArr* image, CvArr*
Edges, double threshold1, double threshold2, int aperture_size=3).Function cvCanny adopts
These edges of mark with the edge of Canny algorithm search input picture and in output image.
Step 2, the result of the computing according to Canny operator, with morphologic method, edge is refined.Use such as
Under
The edge that template elements are more than 3 to width carries out the etching operation in morphology.
Step 3, selects suitable threshold value to carry out the extraction of marginal point and connection in PU.Here, using threshold value ThCarry out side
The extraction of edge point, and choose optimal threshold with iterative method.Iterative method to realize step as follows：
Obtain minima Z in the current PU of imagemin, maximum Zmax, then Initial Hurdle T0=(Zmin+Zmax)/2.
According to threshold values Tk(k is iterationses) divides the image into into marginal area and non-edge, calculates two-part
Meansigma methodss ZO、ZB, that is,
Here, hiAnd hjRepresent the value of the pixel of edge and non-edge respectively.Calculate ZO、ZBAfterwards, use formula
Tk+1=(ZO+ZB)/2
To calculate new threshold values TK+1.
If Tk+1=TK, i.e. TkFor required threshold value, then algorithm leaves it at that, and otherwise goes to step (2).Iterative calculation is until receive
Hold back in certain stable threshold value, this threshold value is final result Th.
Here, detected from candidate marginal using dual threshold method and connect final edge.In the extraction of marginal point,
The threshold value being adopted is Th, the threshold value adopting during the connection of marginal point should compare ThLittle, with adjoining edge.At edge in the present invention
The threshold value adopting during the connection of point is Th/2.That is, after processing through the first and second steps, value is more than ThPixel be marginal point.So
Afterwards with these marginal points as seed, find the value being adjacent and be more than Th/ 2 pixel, adds as edge pixel point.Then, weight
This process multiple, in PU, all of pixel all scans and goes over.
Step 4, to current PU, using Hough transform detection of straight lines edge pixel point, i.e. retains and is on a certain straight line
Marginal point, remove and be in marginal point on curve.Carry out the inspection of straightway by function cvHoughLines2 in OpenCV
Survey.The function call prototype of cvHoughLines2 is CvSeq* cvHoughLines2 (CvArr* image, void* line
Storage, int method, double rho, double theta, int threshold, double param1, double
param2).Parameter method is arranged to CV_HOUGH_PROBABILISTIC and represents select probability Hough transformation by the present invention, enters
The detection of row straightway.Threshold is threshold parameter, if corresponding aggregate-value is more than it, function can return this line segment.
Param1 represents minimum line segment length, and param2 represents the largest interval value connecting in same straight line enterprising line section.
Step 5, to each linear edge, determines the differential seat angle θ in itself and horizontal direction, and its tangent value tan θ.
Step 6, in the scope of the tangent value of the pre- measuring angle according to each intra prediction mode, and the current PU detecting
The tangent value of each linear edge, with statistic histogram determine linear edge pixel that current PU declines in the range of each it
With.To the scope of statistics of the tangent value of linear edge angle as shown in Table 1 and Table 2.For example, as shown in table 1, for candidate pattern
7, the scope of statistics of its angle tangent value be (- 0.7417, -1.1033].That is, the linear pixel point value in the range of initial this is put
0, C7=0, then the tangent value if there are the angle of linear edge fall this scope (- 0.7417, -1.1033] in, then C7=C7
+ai, wherein aiIt is all pixels points on this straight line in current PU.All of candidate pattern is all carried out with this process, then
Obtain the statistic histogram of desired each pattern.Because pattern 0, pattern 3 are non-directional prediction patterns, therefore do not existed by statistics
In table 1, table 2, they are all fixed and enter into roughly selecting of RDO as candidate pattern.
Step 7, according to this statistic histogram, determines several predictions of the linear edge pixel sum maximum in rectangular histogram
Pattern.
Step 8, according to the size of current PU, and the statistic histogram obtaining, determine Dietary behavior rougher process RMD
The frame mode of candidate.Further, the PU to a certain size, the predictive mode of candidate of entrance RDO detection that it selects
Number is as shown in table 6.For example, for the PU of 16 × 16 sizes, 5 will be selected in the maximum candidate of above statistic histogram intermediate value
Pattern.It can be seen that present invention substantially reduces the pattern of the candidate entering into RMD rougher process from the contrast of table 3 and table 4
Number.As such, it is possible to greatly reduce amount of calculation, accelerate the speed of service of HEVC standard.
Step 9, according to RMD process, and the size of current PU, determine the frame in of the candidate of Dietary behavior refining process RDO
Pattern.Further, the PU to a certain size, number such as table 6 institute of the predictive mode of candidate of entrance RDO detection that it selects
Show.For example, for the PU of 16 × 16 sizes, 3 candidate pattern selecting optimum during RMD are entered into final RDO
Selection course in.It can be seen that present invention substantially reduces the time entering into RDO rougher process from the contrast of table 5 and table 6
The number of the pattern of choosing.As such, it is possible to greatly reduce amount of calculation, accelerate the speed of service of HEVC standard.
Step 10, according to the result of RDO, determines the intra prediction mode of the optimization of current PU.
The present invention, according to general test environment, is tested to the method being proposed.Coded format is full I frame knot
Structure, test QP point is 37,32,27,22, using adaptive binary arithmetic coding CABAC (Context-based Adaptive
Binary Arithmetic Coding) entropy code is carried out to video sequence.High score can be widely applied in view of HEVC future
In resolution video sequence, test mainly based on high definition and SD cycle testss, mainly have 25601 × 600p, 1920 ×
1080p, 1280 × 720p and 832 × 480p, the video test sequence of totally 4 kinds of sizes.Here mainly from the new method being proposed
The minimizing of the scramble time bringing and the corresponding cost the paid performance to consider the method.Performance indications are mainly by BD-
The parameters such as PSNR/Rate, Δ B, Δ P and Δ T specifically to express the performance of proposed method.
The experimental results of table 7. method proposed by the invention
Experimental result as shown in table 7, if positive number is then it represents that method of the present invention phase under this index in all data results
It is to increase for its numerical value of former method, if negative is then it represents that under this index, the method for the present invention is with respect to its number of former method
Value is to reduce.The index of test performance is defined as follows：
(1) the scramble time Δ T saving, that is,
Wherein, THM4.0Represent the scramble time using the existing algorithm of HM4.0, TProposedIt is the coding of carried algorithm in literary composition
Time.Under the conditions of obtaining equal code efficiency, Δ T is bigger, and presentation code end computation complexity reduces more, to be proposed methods
Performance better.
(2) method being proposed and HM4.0 intraframe prediction algorithm code check difference DELTA B, that is,
Wherein, BHM4.0Represent the encoder bit rate of existing HM4.0 method, BProposedIt is the coding of method proposed by the invention
Code check.Δ B is less, represent the coding efficiency of the method and the HM4.0 existing method that propose closer to.
(3) method being proposed and Y-PSNR (the Peak Signal to of HM4.0 intra-frame prediction method luminance component
Noise Ratio, PSNR) difference DELTA P, that is,
Δ P=PProposed-PHM4.0
Wherein, PHM4.0Represent the Y-PSNR PSNR, P of the luminance component of HM4.0 existing methodProposedIt is the present invention
The PSNR of the luminance component of proposed method.Δ P is less, represents that the coding efficiency of the method and HM4.0 existing method proposing is got over
Close.
BD-Rate represents under same Y-PSNR (PSNR), the value added of code check.BD-PSNR represents under same code check,
The reduced value of PSNR.
It can be seen that the method being proposed only has the reduction of minimum Y-PSNR (PSNR) from experimental result table 7
And minimum code check increases, therefore it has preferable performance.Meanwhile, method proposed by the invention be can be seen that by this table
The scramble time that minimum 16.15%, maximum can reach 32.73% can be saved with respect to the standard method in HM4.0.
By experimental result obtain to proposed method and former method (HM4.0) to the video measurement sequence under different resolution
The curve of the code check of row and the distortion factor (Rate-Distortion) is as shown in Fig. 3,4,5,6.The abscissa of these figures is code check,
Unit is thousand bit numbers (kbps) that need to transmit per second.The vertical coordinate of these figures is Y-PSNR (PSNR), and unit is dB.
From these in figures it can be seen that two curve substantially phases of the code check of method proposed by the invention and former method and the distortion factor
With.Thus, it is possible to the proposed method of explanation maintains the precision of predictive mode, under conditions of identical code check, do not dislike
Change or than the significantly reduction distortion factor.
It can be seen that the method being proposed, with respect to original method, can averagely be saved from table 7 and Fig. 3,4,5,6
23% scramble time.Meanwhile, the fluctuation of its code check and PSNR is less, for the calculation cost saved, in performance
Reduce and almost can ignore.It can be seen that, method proposed by the invention has preferable performance.
The preferred embodiment of the present invention described in detail above.It should be appreciated that the ordinary skill of this area need not be created
The property made work just can make many modifications and variations according to the design of the present invention.Therefore, all technical staff in the art
Pass through the available technology of logical analysis, reasoning, or a limited experiment under this invention's idea on the basis of existing technology
Scheme, all should be in the protection domain being defined in the patent claims.
Claims (7)
1. a kind of method carrying out quickly determining intra-prediction code mode in international video compression standards HEVC, described side
Method includes：
Step one, to current prediction unit (PU), judges whether its size is 4 × 4, in this way, then carries out the pattern that step-length is 1 sub-
Sampling, and leap to step 8；Otherwise, using Canny operator, row operation is entered to current PU；
Step 2, the result of the computing according to Canny operator, with morphologic method, edge is refined；
Step 3, selects suitable threshold value to carry out the extraction of marginal point and connection in PU；
Step 4, to current PU, using Hough transform detection of straight lines edge pixel point, i.e. retain the side being on a certain straight line
Edge point, removes the marginal point being on curve；
Step 5, to each linear edge, determines the differential seat angle θ in itself and horizontal direction, and its tangent value tan θ；
Step 6 is each straight in the scope of the tangent value of the pre- measuring angle according to each intra prediction mode, and the current PU detecting
The tangent value at line edge, determines, with statistic histogram, the linear edge pixel sum that current PU declines in the range of each；
Step 7, according to this statistic histogram, determines several predictive modes of the linear edge pixel sum maximum in rectangular histogram；
Step 8, according to the size of current PU, and the statistic histogram obtaining, determine the time of Dietary behavior rougher process (RMD)
The frame mode of choosing；
Step 9, according to RMD process, and the size of current PU, determine the frame in mould of the candidate of Dietary behavior refining process (RDO)
Formula；
Step 10, according to the result of RDO, determines the intra prediction mode of the optimization of current PU.
2. the method for claim 1, is more than 3 using morphologic method to the width detecting in described step 2
Edge refined, i.e. carry out morphologic erosion operation, its corrosion template elements be
3. the method for claim 1, adopts threshold value T in described step 3hCarry out marginal point extraction, using iterative method
Choose optimal threshold, it is as follows that iterative method realizes step：
(1) obtain minima Z in the current PU of imagemin, maximum Zmax, then Initial Hurdle T0=(Zmin+Zmax)/ 2,
(2) according to threshold values Tk(k is iterationses) divides the image into into edge pixel point and non-edge pixels point, calculates two parts
Meansigma methodss ZO、ZB, that is,
Here, hiAnd hjRepresent the value of edge pixel point and non-edge pixels point respectively, calculate ZO、ZBAfterwards, use formula
Tk+1=(ZO+ZB)/2
To calculate new threshold values Tk+1；
(3) if Tk+1=Tk, i.e. TkFor required threshold value, then algorithm leaves it at that, and otherwise goes to step (2), and iterative calculation is until receive
Hold back in certain stable threshold value, this threshold value is final result Th.
4. the method for claim 1, in described step 3 in the extraction of marginal point, the threshold value being adopted is Th, side
The threshold value adopting during the connection of edge point should compare ThLittle, with adjoining edge；The threshold adopting in the connection of marginal point in the present invention
It is worth for Th/ 2, i.e. after processing through the first and second steps, value is more than ThPixel be marginal point；Then with these marginal points for planting
Son, finds the value being adjacent and is more than Th/ 2 pixel, adds as edge pixel point；Then, repeat this process, in PU
All of pixel all scans to be gone over.
5. the method for claim 1, the tangent of the pre- measuring angle according to each intra prediction mode in described step 6
The scope of value, and in the current PU detecting each linear edge tangent value, determine that current PU declines every with statistic histogram
Linear edge pixel sum within the scope of one, is moved towards with the edge determining current PU.
6. the method for claim 1, to size 4 × 4,8 × 8,16 × 16,32 × 32 and 64 in described step 8
× 64, select 9,7,5,3,1 candidate pattern to enter into the detection process of RMD respectively；Wherein to PU a size of 8 × 8,16 ×
16th, 32 × 32 and 64 × 64 predicting unit, the candidate pattern selecting statistic histogram cathetus edge pixel point sum big is entered
Enter to RMD detection process, to determine the pattern entering RMD detection process.
7. the method for claim 1, to size 4 × 4,8 × 8,16 × 16,32 × 32 and 64 in described step 9
× 64, select 5,4,3,2,1 candidate pattern to enter into the detection process of RDO respectively；After the standard selecting is RMD process, code
Rate and the little candidate pattern of predicted distortion degree enter into the detection process of RDO, to determine the candidate's mould entering RDO detection process
Formula.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410161592.5A CN103905818B (en) | 2014-04-13 | 2014-04-13 | Method for rapidly determining inter-frame prediction mode in HEVC standard based on Hough conversion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410161592.5A CN103905818B (en) | 2014-04-13 | 2014-04-13 | Method for rapidly determining inter-frame prediction mode in HEVC standard based on Hough conversion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103905818A CN103905818A (en) | 2014-07-02 |
| CN103905818B true CN103905818B (en) | 2017-02-15 |
Family
ID=50996940
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410161592.5A Active CN103905818B (en) | 2014-04-13 | 2014-04-13 | Method for rapidly determining inter-frame prediction mode in HEVC standard based on Hough conversion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103905818B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10070130B2 (en) * | 2015-01-30 | 2018-09-04 | Qualcomm Incorporated | Flexible partitioning of prediction units |
| CN109086743B (en) * | 2018-08-28 | 2022-04-15 | 南京邮电大学 | Lane line detection method based on HEVC compressed video |
| CN112118444B (en) * | 2019-06-20 | 2022-11-25 | 杭州海康威视数字技术股份有限公司 | Encoding method and device |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9232227B2 (en) * | 2011-01-14 | 2016-01-05 | Sony Corporation | Codeword space reduction for intra chroma mode signaling for HEVC |
-
2014
- 2014-04-13 CN CN201410161592.5A patent/CN103905818B/en active Active
Non-Patent Citations (2)
| Title |
|---|
| 一种快速HEVC帧内预测算法;成益龙;《电视技术》;20120525;第36卷(第21期);4-7 * |
| 基于迭代阈值的女书分割算法研究;黑光月;《软件导刊》;20110930;第10卷(第9期);61-63 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103905818A (en) | 2014-07-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109688414B (en) | VVC intra-frame coding unit candidate prediction mode reduction and block division early termination method | |
| Lei et al. | Fast intra prediction based on content property analysis for low complexity HEVC-based screen content coding | |
| Tian et al. | Content adaptive prediction unit size decision algorithm for HEVC intra coding | |
| CN103997646B (en) | Fast intra-mode prediction mode selecting method in a kind of HD video coding | |
| CN103297781B (en) | A kind of HEVC inner frame coding method based on grain direction, device and system | |
| CN105120292A (en) | Video coding intra-frame prediction method based on image texture features | |
| CN103209324B (en) | A kind of method for applying edge compensation | |
| CN104378643B (en) | A kind of 3D video depths image method for choosing frame inner forecast mode and system | |
| CN103517069A (en) | HEVC intra-frame prediction quick mode selection method based on texture analysis | |
| CN107277509A (en) | A kind of fast intra-frame predicting method based on screen content | |
| CN104023233A (en) | Fast inter-frame prediction method of HEVC (High Efficiency Video Coding) | |
| CN104052994A (en) | Method for HEVC intra-frame prediction mode quick decision based on grading self-adaption | |
| CN105791862B (en) | 3 d video encoding depth map internal schema selection method based on fringe complexity | |
| CN102917225A (en) | Method for quickly selecting HEVC (high-efficiency video coding) inframe coding units | |
| CN107864380A (en) | 3D HEVC fast intra-mode prediction decision-making techniques based on DCT | |
| CN106688238A (en) | Improved reference pixel selection and filtering for intra coding of depth map | |
| CN105187826A (en) | Rapid intra-frame mode decision method specific to high efficiency video coding standard | |
| CN103905818B (en) | Method for rapidly determining inter-frame prediction mode in HEVC standard based on Hough conversion | |
| Sun et al. | A fast intra CU size decision algorithm based on canny operator and SVM classifier | |
| CN102934445A (en) | Methods and apparatuses for encoding and decoding image based on segments | |
| CN102396231A (en) | Image-processing device and method | |
| CN114040211A (en) | AVS 3-based intra-frame prediction rapid decision-making method | |
| CN110868593B (en) | Video CU fast partitioning based on regional decision tree | |
| CN110139098B (en) | Decision tree-based intra-frame fast algorithm selection method for high-efficiency video encoder | |
| CN106878754A (en) | A kind of 3D video depths image method for choosing frame inner forecast mode |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |