CN106878754A - A kind of 3D video depths image method for choosing frame inner forecast mode - Google Patents
A kind of 3D video depths image method for choosing frame inner forecast mode Download PDFInfo
- Publication number
- CN106878754A CN106878754A CN201710082487.6A CN201710082487A CN106878754A CN 106878754 A CN106878754 A CN 106878754A CN 201710082487 A CN201710082487 A CN 201710082487A CN 106878754 A CN106878754 A CN 106878754A
- Authority
- CN
- China
- Prior art keywords
- angle point
- mode
- current
- point
- class
- 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.)
- Granted
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
- H04N19/597—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding specially adapted for multi-view video sequence encoding
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N13/00—Stereoscopic video systems; Multi-view video systems; Details thereof
- H04N13/10—Processing, recording or transmission of stereoscopic or multi-view image signals
- H04N13/106—Processing image signals
- H04N13/161—Encoding, multiplexing or demultiplexing different image signal components
-
- 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
- H04N19/11—Selection of coding mode or of prediction mode among a plurality of spatial predictive coding modes
-
- 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
- H04N19/147—Data rate or code amount at the encoder output according to rate distortion criteria
-
- 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
- H04N19/567—Motion estimation based on rate distortion criteria
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention discloses a kind of 3D video depths image method for choosing frame inner forecast mode.Step is as follows：1) for any depth image, the angle point of frame in is found using self-adaptive angular-point detection algorithm, and preserves its position；2) the conventional intra-frame encoding mode CHIMs of HEVC are selected by pattern coarse system of selection RMD to PU blocks, and adds candidate modes list；3) judge whether current PU meets smoothness condition, leapt to 6) if meeting, otherwise continue 4)；4) introduce grader to classify PU, if current PU is divided into C0Class, represents that 5) current PU, for 6) smooth block then skips to, otherwise continues；5) DMM mode computations are carried out, optimal tapered mode is searched out and is split and be added to mode candidate list；6) rate distortion computation is carried out to candidate pattern list, selects optimal predictive mode.The present invention significantly reduces the number of Intra prediction mode selection, the scramble time needed for reducing infra-frame prediction, while ensure that final decoding end synthesizes the video quality at visual angle.
Description
Technical field
The invention belongs to video coding and decoding technology field, particularly a kind of 3D video depths image Intra prediction mode selection
Method.
Background technology
H.265 as the International video coding standard of a new generation, previous generation is inherited H.264, using macro block as coding
Elementary cell.Main framework to it is H.264 similar, mainly comprising moulds such as infra-frame prediction, inter prediction, conversion, quantization, entropy codes
Formula；It is simultaneously H.265 overall to be divided into coding units, three parts of prediction unit and conversion unit.H.265 can be passed under finite bandwidth
Defeated higher-quality video, i.e., H.265 only needing the bandwidth of H.264 half can just transmit the video of phase homogenous quantities.H.265 use
Various ways reduction code check：When improving compression efficiency and coding quality, improving robustness and error recovery capabilities, reduce real-time
Prolong and obtain time delay and Stochastic accessing time delay with channel, reduce complexity etc..
H.264 encoded using the macro block of fixed size, caused the increase of smooth region encoder bit rate on a large scale and thin
The definition of small texture coding is not enough.H.265 coding unit can select different size of coding unit according to feature of image,
The big I of coding unit is from 8x8 to 64x64.During actual coding, pre-encoded video is divided into 64x64 by encoder first
Maximum coding unit (LCU).For each coding unit, encoder determines optimum code list using the dividing mode of quaternary tree
First (CU) is divided.Searchings that optimal CU is divided needs encoder to carry out 85 rate distortion costs of different demarcation mode to calculate, increasing
The complexity for calculating is added.Simultaneously H.265 in order to preferably using the correlation in image particularly texture image space, by frame in
Predictive mode increases to 35 kinds, wherein having for 33 kinds of angle intra prediction modes of texture region prediction and for flat site
DC the and Planar both of which of prediction, as shown in Figure 1.The selection of predictive mode is not with CU as elementary cell but with PU
Carried out as elementary cell.PU is exactly that predicting unit is further to be split to obtain by CU.Intra prediction mode increase so that
Each PU will carry out 35 rate distortion computations and may find optimal predictive mode, which increases the complexity for calculating.
H.265 compared to H.264, an important extension standards are exactly 3D-HEVC.3D-HEVC is conciliate using coding side
The asymmetrical multiple views plus depth picture format in code end.This form can be virtual using the synthesis of DIBR technologies is rendered in decoding end
Viewpoint.Depth image is different with texture image, and its pixel value reaction is distance of the scene to the distance of camera lens.Depth map
Feature is mainly included：1) most flat site；2) sharp keen edge；3) it is not used in direct viewing.The coding of depth image
It is different from Texture Encoding, to retain the edge in figure.Therefore 3D-HEVC retains the side of object using new technology
Edge.Most important of which is that using four kinds of new depth model patterns：Border chain type coding mode (RBC), the wedge shape point of display
Cut pattern (DMM1), implicit wedge-shaped Fractionation regimen (DMM3), contour Fractionation regimens (DMM4).Wedge-shaped Fractionation regimen such as Fig. 2
It is shown, wherein (a) is consecutive hours wedge shape partitioning scheme schematic diagram, wedge-shaped partitioning scheme schematic diagram when (b) is discrete, (c) is final
Wedge-shaped Fractionation regimen, is drawn a straight line by starting point S, terminal E and prediction block PU is divided into two parts.
, when optimal intra prediction mode is selected, intra prediction mode in 35 is not only being calculated during actual coding
Rate distortion costs, will also calculate the rate distortion costs of wedge-shaped Fractionation regimen, and the increase of computation complexity seriously hinders 3D-HEVC
Real-time application.
The content of the invention
It is an object of the invention to provide a kind of 3D video depths image method for choosing frame inner forecast mode, can ensure
In synthesis visual angle on the premise of video quality, reduce depth image intra prediction mode number and reduce depth image infra-frame prediction
The computation complexity of model selection.
The technical solution for realizing the object of the invention is：A kind of 3D video depths image Intra prediction mode selection side
Method, comprises the following steps：
Step 1：For any depth image, the angle point of frame in is found using self-adaptive angular-point detection algorithm, and preserve angle
The position of point；
Step 2：It is that PU blocks select the conventional intra-frame encoding modes of HEVC by the coarse system of selection RMD of pattern to predicting unit
CHIMs, and add candidate modes list；
Step 3：Judge whether current PU meets smoothness condition, current PU is smooth PU if meeting, and leaps to step
Rapid 6, otherwise continue step 4；
Step 4：Introduce grader to classify PU, if current PU is classified device and is divided into C0Class, represents that current PU is flat
Sliding block then skips to step 6, otherwise continues step 5；
Step 5：DMM mode computations are carried out, optimal tapered mode is searched out and is split and be added to mode candidate list；
Step 6：Rate distortion computation is carried out to candidate pattern list, optimal predictive mode is selected.
Further, self-adaptive angular-point detection algorithm is described in step 1：Shi-Tomasi Corner Detection Algorithms；
Two principal directions of angle point are obtained by using principal component analysis in the Shi-Tomasi Corner Detection Algorithms；Square
Two characteristic vectors of battle array M are two principal directions of sampled pixel point, a minimum of two principal direction of each angle point, the feature of matrix
Value is reflected in the intensity of variation of pixel value on the specific direction described by two characteristic vectors.
Further, smoothness condition described in step 3 is what Merkle and M.Zhang et al. were proposed：The variance of current PU is small
In a threshold value, and RMD judges that Planar pattern complexities are minimum and this PU is smooth PU during with minimum rate distortion costs.
Further, grader is introduced described in step 4 to classify PU, it is specific as follows：
(4.1) the corresponding Metzler matrix of each angle point has two characteristic values, with matrix M small characteristic value come the angle point to candidate
Classified：The point that the absolute value of characteristic value is not more than 0.05 is filtered, by remaining point according to angle point Metzler matrix small feature value
Descending is arranged, and according to the amplitude magnitude classification of characteristic value, now the angle point sum per two field picture is NC；
(4.2) angle point is filtered again with quantization parameter QP, angle point is no longer filtered if the QP of image is not more than 36, pass through
Angle point sum Th after second wheel threshold filterNumAngle point sum N filtered equal to the first roundC, after otherwise filtering the first round
Angle point sum multiply a ratio multiplier determined by QP, the ratio multiplier exists and less than 1 when QP is more than 36, now ThNum
It is NCWith the product of ratio multiplier, last filter result is to find out preceding Th in the first round filtered classification chartNumIndividual angle point.
(4.3) if current PU is free of angle point, it is seen as the smooth PU or continuous PU of direction change；Take NCIn own
Minimal eigenvalue is used as categorised demarcation line in the Metzler matrix characteristic value of angle point, if in two characteristic values of the corresponding Metzler matrix of current PU
Small characteristic value is classified as C less than categorised demarcation line0Class, is otherwise classified as C1Class；C0Class refers to that current PU, for smooth block, can skip DMM
Calculate；C1Class refers to that current PU not can skip DMM calculating.
Compared with prior art, its remarkable advantage is the present invention：(1) angle point information of depth image is make use of to reduce deeply
The intra prediction mode number of image is spent, angle point grader is introduced, by PU according to whether being classified comprising angle point and to difference
The PU of classification carries out different treatment respectively, while ensureing that the marginal information of depth image is not lost；(2) depth is taken full advantage of
The smoothness condition of image PU blocks, the PU blocks that will meet smoothness condition skip angle point classification and tapered mode selection；(3) depth is utilized
The characteristic of image PU is classified, and different skipping methods is used to different classes of PU.Improve the same of the speed of coding
When can be effectively retained border, hereby it is ensured that the video quality at synthesis visual angle.
Brief description of the drawings
Fig. 1 is 35 kinds of intra prediction mode schematic diagrames of HEVC described in background technology.
Fig. 2 is the partitioning scheme of wedge-shaped Fractionation regimen described in background technology, wherein (a) is consecutive hours wedge shape partitioning scheme
Schematic diagram, wedge-shaped partitioning scheme schematic diagram when (b) is discrete, (c) is final wedge-shaped Fractionation regimen.
Fig. 3 is the structure chart of 3D video depths image method for choosing frame inner forecast mode of the present invention.
Fig. 4 is the general frame figure of 3D video depths image method for choosing frame inner forecast mode of the present invention.
Fig. 5 is the angle point explanatory diagram in embodiment, wherein (a) is that window is in all directions when smooth region is moved
Variation diagram, (b) is the window variation diagram in all directions when corner point is moved, and (c) is that window exists when being moved on edge
The variation diagram of edge direction.
Fig. 6 is the PU classification charts in 3D video depths image method for choosing frame inner forecast mode of the present invention.
Specific embodiment
3D video depths image method for choosing frame inner forecast mode proposed by the present invention, by extracting depth image angle point pair
Prediction block PU is classified, to C0The PU blocks of class directly skip DMM model selections, and then reduction needs to calculate rate distortion costs
Intra prediction mode number.As shown in figs. 34,3D video depths image method for choosing frame inner forecast mode of the present invention, including such as
Lower step：
Step 401：For any depth image, the angle point of frame in is found using self-adaptive angular-point detection algorithm, and preserved
The position of angle point；
Step 402：It is that PU blocks select the conventional intraframe coding moulds of HEVC by the coarse system of selection RMD of pattern to predicting unit
Formula CHIMs, and add candidate modes list；
Step 403：Judge whether current PU meets smoothness condition, current PU is smooth PU if meeting, and is leapt to
Step 406, otherwise continues step 404；
Step 404：Introduce grader to classify PU, if current PU is classified device and is divided into C0Class, represents that current PU is
Smooth block then skips to step 6, otherwise continues step 5；
Introduce angle point grader to classify PU, if current PU is classified device and is divided into C0Class, then skip to step 406, no
Then, step 405 is continued.
Step 405：DMM mode computations are carried out, optimal tapered mode is searched out and is split and be added to mode candidate list；
Step 406：Rate distortion computation is carried out to candidate pattern list, optimal predictive mode is selected.
The general idea that self-adaptive angular-point described in above-mentioned steps 401 is extracted is as follows：
When window is moved on image, when smooth region is moved such as shown in Fig. 5 (a), window does not all have in all directions
Change；When being moved on edge such as Fig. 5 (c), window is not changed in the direction at edge；When corner point is moved such as Fig. 5
B (), window is changed in all directions.Using this intuitively physical phenomenon, by window change in all directions
Change degree, determines whether current point is angle point.
The definition of angle point is：Two intersection points on side, tightened up says, the local neighborhood of angle point should have two not same districts
The border of the different directions in domain.
Self-adaptive angular-point detection algorithm is described in above-mentioned steps 401：Shi-Tomasi Corner Detection Algorithms；The Shi-
Two principal directions of angle point are obtained by using principal component analysis in Tomasi Corner Detection Algorithms；Two features of matrix M to
Amount is two principal directions of sampled pixel point, and a minimum of two principal direction of each angle point, the characteristic value of matrix is reflected in two spies
Levy the intensity of variation of pixel value on the specific direction described by vector.
Smoothness condition described in above-mentioned steps 3 is what Merkle and M.Zhang et al. were proposed：The variance of current PU is less than one
Threshold value, and RMD judge Planar pattern complexities it is minimum and during with minimum rate distortion costs this PU in smooth PU, i.e. Fig. 6
PU1。
Grader is introduced described in above-mentioned steps 4 to classify PU, it is specific as follows：
(4.1) the corresponding Metzler matrix of each angle point has two characteristic values, with matrix M small characteristic value come the angle point to candidate
Classified：The point that the absolute value of characteristic value is not more than 0.05 is filtered, by remaining point according to angle point Metzler matrix small feature value
Descending is arranged, and according to the amplitude magnitude classification of characteristic value, now the angle point sum per two field picture is NC；
(4.2) angle point is filtered again with quantization parameter QP, angle point is no longer filtered if the QP of image is not more than 36, pass through
Angle point sum Th after second wheel threshold filterNumAngle point sum N filtered equal to the first roundC, after otherwise filtering the first round
Angle point sum multiply a ratio multiplier determined by QP, the ratio multiplier exists and less than 1 when QP is more than 36, now ThNum
It is NCWith the product of ratio multiplier, last filter result is to find out preceding Th in the first round filtered classification chartNumIndividual angle point；
(4.3) if current PU is free of angle point, it is seen as the smooth PU or continuous PU of direction change；Take NCIn own
Minimal eigenvalue is used as categorised demarcation line in the Metzler matrix characteristic value of angle point, if in two characteristic values of the corresponding Metzler matrix of current PU
Small characteristic value is classified as C less than categorised demarcation line0Class, is otherwise classified as C1Class；C0Class refers to that current PU, for smooth block, can skip DMM
Calculate；C1Class refers to that current PU not can skip DMM calculating.C0Class includes the PU in Fig. 62To PU5。
Embodiment 1
In order that the object, technical solutions and advantages of the present invention become more apparent, with reference to embodiment, to the present invention
Further describe.
The embodiment of the present invention is to the performance of the 3D video depth image intra prediction mode skipping methods of proposition in 3D-HEVC
Reference software HTM-13.0 in verified, use conventional testing conditions (CTC) using coding parameter：
Video sequence resolution ratio：1920x1088、1024x768.
Test frame type：Full I frames
Depth model pattern (DMM)：Open
Quantization parameter value：Texture image：25 34 30 39 depth images：35 42 40 45
Simplify depth coding (SDC)：Open
Loop filtering：Close
Sample adaptive equalization filters SAO：Close
Visual angle synthesis optimizing VSO：Open
The present embodiment operating process is as shown in figure 4, comprise the following steps that：
Step 401：For any depth image, find the angle point of frame in using self-adaptive angular-point detection algorithm and preserve it
Positional information.
Step 402：Common HEVC intra prediction modes are selected by RMD to PU blocks, and adds candidate modes list
In.
Step 403：Whether the variance of current PU is judged less than a threshold value, and RMD judges that Planar pattern complexities are
It is no minimum i.e. with minimum rate distortion costs, PU during current PU is divided into smooth PU i.e. Fig. 6 if meeting1, leap to step
Rapid 406, otherwise continue step 404.
Step 404：Introduce angle point grader to classify PU, if current PU is classified device and is divided into C0Class, then skip to step
Rapid 406, otherwise, continue step 405.
Step 405：DMM model selections are carried out, optimal wedge shape segmentation is searched out and is added to mode candidate list.
Step 406：Rate distortion computation is carried out to candidate pattern list, optimal predictive mode is selected.
Encoding efficiency is evaluated using code check (BR) and Y-PSNR (PSNR).Δ T represented and encoded compared with HTM methods
The saving of time.
The angle point grader division result table of table 1
As can be seen from Table 1 when PU is divided into C0During class, these different sizes PU selects DMM patterns as optimum segmentation pattern
Average probability be respectively 1.62%, 0.69%, 0.38% and 0.48%；Meanwhile, there is 24.94%, 19.50%, 14.01% and
11.01% is divided into C1The PU of class selects DMM as optimum segmentation pattern.This shows nearly all to be divided into C0The PU of class can
Predicted well by conventional HEVC intra-frame encoding modes (CHIMs).Therefore, these PU skip DMM and calculate and can ensure to encode
Reduce while quality time-consuming.
The method provided by the present invention result of table 2 and HTM13.0 method comparative result tables
Depth image intra-frame encoding mode skip algorithm (DMMSA) experimental result that the present invention is provided, as shown in table 2,
DMMSA algorithms can save about 17% when consumption compared with the method that 3D-HEVC is used, averagely.DMMSA algorithms are comprising more flat
Effect is more preferable in Dancer, Fly, Hall2 of skating area domain and simple PU.Other this algorithm only brings the increasing of 0.19% code check
Plus, the loss very little of this explanation video quality compared with HTM-13.0.But the code check of video sequence Dancer, Hall2 is slightly higher.
In these video sequences, depth image pixel value amplitude of variation is small and change is slow, in an encoding process using DMMSA algorithms
Skip DMM mode computations.But, the color video of corresponding region is more complicated, causes the decline of composograph quality.Fly videos
The region but its code check that sequence equally includes pixel value gradual change only have 0.03% increase, because its correspondence texture image is
One flat background (desert).
The present invention improves the efficiency of depth image intraframe coding, companion while saving the depth image intraframe coding time
With being slightly increased for code check.In the case of quantization parameter is less, Dancer, Hall2, Fly video sequence are carried using the present invention
The time of intraframe coding during the method for confession is saved more than 25%, and with the time for the increasing saving reduction of quantization parameter.This is
Because the threshold value for using in step 303 is relevant with quantization parameter, and the grader based on angle point for using in step 304
It is also relevant with quantization parameter.The angle point for extracting is screened when quantization parameter is more than 36.These three video sequences
Row include most smooth region, and only 10%~20% PU needs to test DMM as threshold value with big quantization parameter
Pattern, the threshold method for now being used in step 303 has played Main Function.But when quantization parameter is smaller, use small threshold
Value ensures picture quality so that PU changes somewhat can all influence the DMM model selections of PU.Due to multidirectional feature angle point
PU slight changes can be solved the problems, such as so that PU small changes do not influence DMM model selections, so the present invention is proposed
The sorting technique based on angle point be better than HTM methods.DMMSA methods can be by the adaptively selected quantization parameter of angle point, will more
Many PU points is C0Class is calculated so as to skip DMM.
The angle point of table 3 calculates timetable
As shown in table 3, the detection time of angle point only has 0.301s, accounts for the 0.6% of depth image intraframe coding time.Therefore
The calculating of angle point does not interfere with the encoding efficiency of DMMSA algorithms.
The angle point scheme performance table of table 4
Table 4 is three kinds of angle point Choices,WithScheme exactly subtracts when QP increases by 3 angle point quantity
Few half.WithRepresent respectively and reduce 1/3rd and 2/3rds when QP often increases by 3 hour angle point quantity.In general,Scheme can obtain preferable result, so selection schemeIt is applied to grader.
Claims (4)
1. a kind of 3D video depths image method for choosing frame inner forecast mode, it is characterised in that comprise the following steps：
Step 1：For any depth image, the angle point of frame in is found using self-adaptive angular-point detection algorithm, and preserve angle point
Position；
Step 2：It is that PU blocks select the conventional intra-frame encoding modes of HEVC by the coarse system of selection RMD of pattern to predicting unit
CHIMs, and add candidate modes list；
Step 3：Judge whether current PU meets smoothness condition, current PU is smooth PU if meeting, and leaps to step 6,
Otherwise continue step 4；
Step 4：Introduce grader to classify PU, if current PU is classified device and is divided into C0Class, represents current PU for smooth block then
Step 6 is skipped to, otherwise continues step 5；
Step 5：DMM mode computations are carried out, optimal tapered mode is searched out and is split and be added to mode candidate list；
Step 6：Rate distortion computation is carried out to candidate pattern list, optimal predictive mode is selected.
2. 3D video depths image method for choosing frame inner forecast mode as claimed in claim 1, it is characterised in that step 1 institute
Self-adaptive angular-point detection algorithm is stated for Shi-Tomasi Corner Detection Algorithms；
Two principal directions of angle point are obtained by using principal component analysis in the Shi-Tomasi Corner Detection Algorithms；Matrix M
Two characteristic vectors be sampled pixel point two principal directions, a minimum of two principal direction of each angle point, the characteristic value of matrix
It is reflected in the intensity of variation of pixel value on the specific direction described by two characteristic vectors.
3. 3D video depths image method for choosing frame inner forecast mode as claimed in claim 1, it is characterised in that step 3 institute
It is Merkle and M.Zhang et al. proposition to state smoothness condition：The variance of current PU is less than a threshold value, and RMD judges
Planar pattern complexities are minimum and this PU is smooth PU during with minimum rate distortion costs.
4. 3D video depths image method for choosing frame inner forecast mode as claimed in claim 1, it is characterised in that step 4 institute
Introducing grader is stated to classify PU, it is specific as follows：
(4.1) the corresponding Metzler matrix of each angle point has two characteristic values, is carried out come the angle point to candidate with matrix M small characteristic value
Classification：The point that the absolute value of characteristic value is not more than 0.05 is filtered, by remaining point according to angle point Metzler matrix small feature value descending
Arrangement, and according to the amplitude magnitude classification of characteristic value, now the angle point sum per two field picture is NC；
(4.2) angle point is filtered again with quantization parameter QP, angle point is no longer filtered if the QP of image is not more than 36, by second
Angle point sum Th after wheel threshold filterNumAngle point sum N filtered equal to the first roundC, otherwise by the first round filtered angle
Point sum multiplies a ratio multiplier determined by QP, and the ratio multiplier exists and less than 1 when QP is more than 36, now ThNumIt is NC
With the product of ratio multiplier, last filter result is to find out preceding Th in the first round filtered classification chartNumIndividual angle point；
(4.3) if current PU is free of angle point, it is seen as the smooth PU or continuous PU of direction change；Take NCIn all angle points
Metzler matrix characteristic value in minimal eigenvalue as categorised demarcation line, if small in two characteristic values of the current corresponding Metzler matrix of PU
Characteristic value is classified as C less than categorised demarcation line0Class, is otherwise classified as C1Class；C0Class refers to current PU for smooth block, can skip DMM meters
Calculate；C1Class refers to that current PU not can skip DMM calculating.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710082487.6A CN106878754B (en) | 2017-02-16 | 2017-02-16 | A kind of 3D video depth image method for choosing frame inner forecast mode |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710082487.6A CN106878754B (en) | 2017-02-16 | 2017-02-16 | A kind of 3D video depth image method for choosing frame inner forecast mode |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106878754A true CN106878754A (en) | 2017-06-20 |
| CN106878754B CN106878754B (en) | 2019-09-27 |
Family
ID=59166068
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710082487.6A Expired - Fee Related CN106878754B (en) | 2017-02-16 | 2017-02-16 | A kind of 3D video depth image method for choosing frame inner forecast mode |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106878754B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108449591A (en) * | 2018-03-21 | 2018-08-24 | 天津大学 | A kind of deep video DMM patterns simplification method |
| CN112235570A (en) * | 2020-09-28 | 2021-01-15 | 中南大学 | Fast prediction method based on precoding |
| CN113613006A (en) * | 2021-07-30 | 2021-11-05 | 浙江裕瀚科技有限公司 | Method, system and device for video coding |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104378643A (en) * | 2014-12-04 | 2015-02-25 | 南京理工大学 | Intra-frame prediction mode selection method and system of 3D (3-dimension) video plus depth image |
| CN105007494A (en) * | 2015-07-20 | 2015-10-28 | 南京理工大学 | Intra-frame wedge-shaped segmentation mode selection method of 3D video depth image model mode |
-
2017
- 2017-02-16 CN CN201710082487.6A patent/CN106878754B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104378643A (en) * | 2014-12-04 | 2015-02-25 | 南京理工大学 | Intra-frame prediction mode selection method and system of 3D (3-dimension) video plus depth image |
| CN105007494A (en) * | 2015-07-20 | 2015-10-28 | 南京理工大学 | Intra-frame wedge-shaped segmentation mode selection method of 3D video depth image model mode |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108449591A (en) * | 2018-03-21 | 2018-08-24 | 天津大学 | A kind of deep video DMM patterns simplification method |
| CN112235570A (en) * | 2020-09-28 | 2021-01-15 | 中南大学 | Fast prediction method based on precoding |
| CN112235570B (en) * | 2020-09-28 | 2024-04-02 | 中南大学 | Depth prediction method based on precoding and intra-frame direction prediction method |
| CN113613006A (en) * | 2021-07-30 | 2021-11-05 | 浙江裕瀚科技有限公司 | Method, system and device for video coding |
| CN113613006B (en) * | 2021-07-30 | 2023-08-18 | 浙江裕瀚科技有限公司 | Video coding method, system and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106878754B (en) | 2019-09-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111147867B (en) | Multifunctional video coding CU partition rapid decision-making method and storage medium | |
| CN109688414B (en) | VVC intra-frame coding unit candidate prediction mode reduction and block division early termination method | |
| CN110087087B (en) | VVC inter-frame coding unit prediction mode early decision and block division early termination method | |
| CN103997646B (en) | Fast intra-mode prediction mode selecting method in a kind of HD video coding | |
| EP3389276B1 (en) | Hash-based encoder decisions for video coding | |
| CN101507277B (en) | Image encoding/decoding method and apparatus | |
| CN109302610B (en) | Fast coding method for screen content coding interframe based on rate distortion cost | |
| EP3962080A1 (en) | Method and apparatus for image processing | |
| Shen et al. | Ultra fast H. 264/AVC to HEVC transcoder | |
| CN104125466B (en) | A kind of HEVC parallel decoding methods based on GPU | |
| Jamali et al. | Fast HEVC intra mode decision based on edge detection and SATD costs classification | |
| WO2013074365A1 (en) | Subjective based post-filter optimization | |
| CN102934446A (en) | Determining intra prediction mode of image coding unit and image decoding unit | |
| CN109040764B (en) | HEVC screen content intra-frame rapid coding algorithm based on decision tree | |
| CN101710993A (en) | Block-based self-adaptive super-resolution video processing method and system | |
| CN106170093B (en) | Intra-frame prediction performance improving coding method | |
| US20150195519A1 (en) | Video encoder with intra-prediction candidate screening and methods for use therewith | |
| CN105208387A (en) | HEVC intra-frame prediction mode fast selection method | |
| CN107018412B (en) | A kind of DVC-HEVC video transcoding method based on key frame coding unit partition mode | |
| CN106878754B (en) | A kind of 3D video depth image method for choosing frame inner forecast mode | |
| CN109743575A (en) | A kind of DVC-HEVC video transcoding method based on naive Bayesian | |
| CN103313055B (en) | A kind of chroma intra prediction method based on segmentation and video code and decode method | |
| CN102934445A (en) | Methods and apparatuses for encoding and decoding image based on segments | |
| CN113489976A (en) | Predictive coding method and device and electronic equipment | |
| CN109889838B (en) | HEVC (high efficiency video coding) rapid coding method based on ROI (region of interest) |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | 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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190927 Termination date: 20210216 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |