CN108271025A - The coding circuit of depth modelling pattern and its coding method in 3D coding and decoding videos based on boundary gradient - Google Patents
The coding circuit of depth modelling pattern and its coding method in 3D coding and decoding videos based on boundary gradient Download PDFInfo
- Publication number
- CN108271025A CN108271025A CN201810060934.2A CN201810060934A CN108271025A CN 108271025 A CN108271025 A CN 108271025A CN 201810060934 A CN201810060934 A CN 201810060934A CN 108271025 A CN108271025 A CN 108271025A
- Authority
- CN
- China
- Prior art keywords
- grad
- block
- search
- coarse search
- value
- 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
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
- H04N19/17—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object
- H04N19/176—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object the region being a block, e.g. a macroblock
-
- 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
- H04N19/119—Adaptive subdivision aspects, e.g. subdivision of a picture into rectangular or non-rectangular coding blocks
-
- 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
- H04N19/42—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation
- H04N19/436—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation using parallelised computational arrangements
-
- 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
- H04N19/593—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving spatial prediction techniques
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Computing Systems (AREA)
- Theoretical Computer Science (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
Coding circuit and its coding method the invention discloses depth modelling pattern in a kind of 3D coding and decoding videos based on boundary gradient, including input data processing module, coarse search module and smart search module；Input data processing module is the channel that external reference data are inputted into module, and gradient data is handled；Coarse search module is used to carry out initial forecast to depth map；Smart search module is used to carry out essence prediction again to the prediction result that coarse search obtains.The present invention can reduce the scramble time, reduce the calculation amount of circuit, and reduce the power consumption of circuit, so as to promote the performance of entire video coding circuit.
Description
Technical field
The invention belongs to the intraframe predictive coding technical fields of video coding and decoding technology, specifically a kind of to be based on boundary
The pattern-coding circuit of depth modelling and its coding method in the 3D coding and decoding videos of gradient.
Background technology
With the continuous development of science and technology, information technology and computer internet are changing people in various degree respectively
Daily life.Nowadays, people obtain information and are mainly derived from multimedia messages, and multimedia messages are using video as core
The heart.3D videos are capable of providing the really and naturally effect that reproduces to user's scene due to comparing common 2D videos and receive industrial quarters
With the attention of academia, become one of hot spot in field of video research.Compared to common 2D videos, 3D includes more huge
Big data volume, this all brings certain difficulty to the transmission and preservation of video data.Therefore, 3D videos are carried out effective
Compressed encoding just shows particularly significant.
In the intra prediction of the depth image of 3D videos, Planar, DC, angle prediction three compared to common 2D videos
Big Predicting Technique, adds DMM depth modelling patterns, and DMM can preferably retain the marginal information of depth image, but same with this
When, result in encoder complexity sharp increase.While ensureing to synthesize viewpoint quality, the high complexity for how reducing DMM is calculated in advance
Method is into an important research direction.It is divided into DMM depth modelling patterns as DMM1, DMM4 both of which.
In the prior art, also it is not very much, Gustavo for the hardware circuit design of DMM depth modelling patterns
Sanchez et al. is 2,017 30thSymposium on Integrated Circuits and Systems Design meetings
On " the Low-Area Scalable Hardware Architecture for DMM-1 Encoder of 3D-HEVC that deliver
Circuit described in Video Coding Standard ", although reducing the circuit of essence search part, also, in each Grad
Parallel work-flow is carried out in the inter-process of directional information, but it remains as string on for different Grad directional informations
Row coding, causes the scramble time long, the circuit computing period is longer, influences coding circuit whole work efficiency in this way.
Invention content
The present invention proposes a kind of 3D videos based on boundary gradient to solve above-mentioned the shortcomings of the prior art
The coding circuit of depth modelling pattern and its coding method in encoding and decoding to reduce the scramble time, shorten the operation of circuit
Period, so as to promote the performance of entire video coding circuit.
To achieve the above object of the invention, the present invention adopts the following technical scheme that：
The coding circuit of depth modelling pattern, remembers any depth in a kind of 3D coding and decoding videos based on boundary gradient of the present invention
The pixel value in 4N × 4N regions in image is spent for original block RU, and wherein N is positive integer, 1≤N≤8；Its main feature is that：The volume
Code circuit includes：Input data processing module, coarse search module, smart search module and wedge block model data store module；
The input data processing module receives externally input original block RU, and calculate the original block RU it is upper and lower,
Left and right four borderline 4N-1 Grad, so as to which one is obtained 4 × (4N-1) a Grad on four boundaries, by institute
There is Grad to be all added, obtain Grad and Sum, then the binary expression mode of the Grad and Sum are moved to right (3+N)
Position, obtains gradient pre-filtering threshold value Thr；Judge to whether there is the ladder less than the gradient pre-filtering threshold value Thr on four boundaries
Angle value, if in the presence of the Grad whole reset of the gradient pre-filtering threshold value Thr will be less than on four boundaries；Again to i-th
Borderline 4N-1 Grad is successively into line label；Described i-th borderline 4N-1 Grad is subjected to descending sort,
So as to obtain the corresponding sequence that Grad corresponds to label, it is denoted as i-th of Grad location information Posi, wherein, Grad is " 0 "
Label be also denoted as " 0 ", 1≤i≤4；With i-th of Grad location information PosiWith j-th of Grad location information PosjAs
One group of Grad directional information, so as to obtain 6 groups of Grad directional informations, wherein any one group of Grad directional information is denoted as
Orit, 1≤t≤6；
The coarse search module is according to t group Grad directional informations OritIn Grad point corresponding to non-" 0 " label
It is other that the original block RU is split, so as to obtain KtA coarse search divides block, wherein, k-th of coarse search segmentation block be by
The two wedge blocks composition obtained after the original block RU segmentations；Block is divided according to k-th of coarse search, in the original block RU
Pixel value carry out mean value computation respectively according to two wedge blocks, obtain the mean value of two wedge blocks and be filled into corresponding wedge shape
In block, so as to form k-th of coarse search prediction block；The coarse search rate distortion costs value of k-th of coarse search prediction block is calculated, so as to
Obtain KtThe coarse search rate distortion costs value of a coarse search prediction block；It is thick corresponding to 6 groups of Grad directional informations of synchronous calculating
Rate distortion costs value is searched for, and therefrom selects the coarse search prediction block corresponding to minimum coarse search rate distortion costs value as most
Excellent coarse search prediction block, i ≠ j, 1≤j≤4；
Two Grad of the essence search module according to optimal coarse search prediction block on split position, find phase respectively
Two adjacent Grad, so as to according to a Grad and its adjacent two Grad and another Grad and its adjacent
Two Grad are respectively split the optimal coarse search prediction block, obtain 8 essence search segmentation blocks；Wherein, s-th of essence
Two wedge blocks that search segmentation block obtains after being divided by the original block RU form；It is right according to s-th of essence search segmentation block
Pixel value in the original block RU carries out mean value computation respectively according to two wedge blocks, obtains the mean value of two wedge blocks and fills out
It is charged in corresponding wedge block, so as to form s-th of essence search prediction block；Calculate the smart searching rate of s-th of essence search prediction block
Distortion cost value, so as to obtain the essence search rate distortion costs value of 8 essence search prediction blocks, from this 8 smart searching rate distortion generations
The prediction block corresponding to minimum rate distortion costs value is selected in value and minimum coarse search rate distortion costs value as optimal essence
Prediction block is searched for, residual block is finally calculated with the original block RU according to the optimal essence search prediction block, so as to institute
It states optimal essence search prediction block and residual block realizes that the data compression to the original block RU is transmitted；1≤s≤8.
It is the characteristics of the coding method of depth modelling pattern in a kind of 3D coding and decoding videos based on boundary gradient of the present invention
It carries out as follows：
Step 1,4N × 4N regions in any depth image of note pixel value be original block RU, wherein N is positive integer, 1
≤N≤8；The borderline 4N-1 Grad in four, upper and lower, left and right of the original block RU is calculated, so as on four boundaries
One is obtained 4 × (4N-1) a Grad；
All Grad are all added by step 2, obtain Grad and Sum, then by the two of the Grad and Sum into
Expression way processed moves to right (3+N) position, obtains gradient pre-filtering threshold value Thr；
Judge to whether there is the Grad less than gradient pre-filtering threshold value Thr on four boundaries, if in the presence of by four sides
It is less than the Grad whole reset of gradient pre-filtering threshold value Thr in boundary；
Step 3,4N-1 Grad borderline to described i-th are and borderline to i-th successively into line label
4N-1 Grad carries out descending sort, obtains the corresponding sequence that Grad corresponds to label, is denoted as i-th of Grad location information
Posi, wherein, Grad is also denoted as " 0 ", 1≤i≤4 for the label of " 0 "；
Step 4, with i-th of Grad location information PosiWith j-th of Grad location information PosjAs one group of Grad
Directional information, so as to obtain 6 groups of Grad directional informations, wherein any one group of Grad directional information is denoted as Orit, 1≤t≤
6；I ≠ j, 1≤j≤4；
Step 5, initialization t=1；
Step 6, according to t group Grad directional informations OritIn Grad corresponding to non-" 0 " label respectively to described
Original block RU is split, so as to obtain KtA coarse search divides block, wherein, k-th of coarse search segmentation block is by described original
The two wedge blocks composition obtained after block RU segmentations；
Step 7 divides block according to k-th of coarse search, to the pixel value in the original block RU according to two wedge blocks point
Mean value computation is not carried out, obtain the mean value of two wedge blocks and is filled into corresponding wedge block, so as to form k-th of coarse search
Prediction block；
Step 8, the coarse search rate distortion costs value for calculating k-th of coarse search prediction block, so as to obtain KtA coarse search is pre-
Survey the coarse search rate distortion costs value of block；
T+1 is assigned to t by step 9, judges whether t ＞ 6 are true, if so, then perform step 9；Otherwise return to step 5；
Step 10 is selected from all coarse search rate distortion costs values corresponding to minimum coarse search rate distortion costs value
Coarse search prediction block is as optimal coarse search prediction block；
Step 11, two Grad according to optimal coarse search prediction block on split position, find adjacent two respectively
A Grad；
Step 12, according to a Grad and its adjacent two Grad and another Grad and its adjacent two
Grad is respectively split the optimal coarse search prediction block, obtains 8 essence search segmentation blocks；Wherein, s-th of essence search
Two wedge blocks that segmentation block obtains after being divided by the original block RU form；
Step 13 divides block according to s-th of coarse search, to the pixel value in the original block RU according to two wedge blocks point
Mean value computation is not carried out, obtain the mean value of two wedge blocks and is filled into corresponding wedge block, so as to form s-th of essence search
Prediction block；
Step 14, the essence search rate distortion costs value for calculating s-th of essence search prediction block, it is pre- so as to obtain 8 essence search
The essence search rate distortion costs value of block is surveyed, from this 8 essence search rate distortion costs values and minimum coarse search rate distortion costs value
The prediction block corresponding to minimum rate distortion costs value is selected as optimal essence search prediction block；
Residual block is calculated with the original block RU according to the optimal essence search prediction block in step 15, so as to institute
It states optimal essence search prediction block and residual block realizes that the data compression to the original block RU is transmitted；1≤s≤8.
Compared with prior art, advantageous effects of the invention are embodied in：
1st, the existing depth image intraframe predictive coding circuit of optimization proposed by the present invention, overcomes coarse search in original design
The problem of process code overlong time, it is proposed that a kind of full parellel circuit framework of coarse search module, to all coarse searches wedge shape
Block splitting scheme is carried out at the same time calculating, the time required to reducing coding.
2nd, the existing depth image intraframe predictive coding circuit of optimization proposed by the present invention, it is first in input data processing module
Pre-filtering first is carried out according to the threshold value of setting to each borderline Grad being calculated, the Grad less than threshold value is all put
" 0 ", in coarse search module later, the location of Grad for " 0 " will be no longer split, and reduce the meter of circuit
Calculation amount, while reduce the power consumption of circuit.
3rd, the existing depth image intra-frame predictive encoding method of optimization proposed by the present invention, changes in original depth image frame
Predict the mode of serial code in DMM1 patterns, it is proposed that a kind of intraframe predictive coding algorithm of full parellel reduces coding and calculates
In the period needed for method operation, save the scramble time.
Description of the drawings
Fig. 1 is the pixel value schematic diagram of original block RU in the prior art；
Fig. 2 is coding circuit general frame figure of the present invention；
Fig. 3 is the Grad pre-filtering schematic diagram of the present invention；
Fig. 4 is Grad location information schematic diagram of the present invention；
Fig. 5 divides block schematic diagram for coarse search of the present invention；
Fig. 6 is coarse search prediction block schematic diagram of the present invention；
Fig. 7 calculates schematic diagram for coarse search rate distortion costs value of the present invention；
Fig. 8 is present invention essence search segmentation block schematic diagram；
Fig. 9 searches for prediction block schematic diagram for present invention essence；
Figure 10 calculates schematic diagram for residual block of the present invention；
Specific embodiment
In the present embodiment, the pixel value for remembering 4N × 4N regions in any depth image is original block RU, and wherein N is just whole
Number, 1≤N≤8；In the present embodiment, N=1, the i.e. original block are the region of one 4 × 4, specific pixel value such as Fig. 1 institutes
Show；
Include as shown in Fig. 2, being somebody's turn to do the coding circuit based on depth modelling pattern in the 3D coding and decoding videos of boundary gradient：It is defeated
Enter data processing module, coarse search module, smart search module and wedge block model data store module；
Input data processing module receives externally input original block RU, and calculates the upper and lower, left and right four of original block RU
A borderline 4N-1 Grad, so as to which one is obtained 4 × (4N-1) a Grad on four boundaries, by all Grad
It is all added, obtains Grad and Sum, then the binary expression mode of Grad and Sum is moved to right into (3+N) position, obtain gradient
Pre-filtering threshold value Thr；Whether four borderline Grad are judged all less than gradient pre-filtering threshold value Thr, if so, not right
Four borderline Grad make any change, otherwise, the Grad that gradient pre-filtering threshold value Thr is less than on four boundaries is complete
Portion's reset, without modification, in the present embodiment, one shares 12 to the Grad more than gradient pre-filtering threshold value Thr on four boundaries
It is all added, obtains Grad and Sum=2+1+45+1+0+1+44+0+0+2+1+1=98 by a Grad, then by gradient
Value and the binary expression mode 1100010 of Sum move to right (3+1)=4, obtain the binary expression side of gradient pre-filtering threshold value
Formula Thr=110, i.e. Thr=6, it is clear that four borderline Grad are not all of being less than gradient pre-filtering threshold value Thr, therefore will
On four boundaries be less than 6 Grad whole reset, the Grad more than 6 without modification, as shown in figure 3, four are borderline
The 44 of 45 and lower boundary of the only remaining coboundary of non-" 0 " Grad；To i-th, borderline 4N-1 Grad carries out successively again
Label；I-th of borderline 4N-1 Grad is subjected to descending sort, so as to obtain the respective row that Grad corresponds to label
Sequence is denoted as i-th of Grad location information Posi, wherein, Grad is also denoted as " 0 ", 1≤i≤4 for the label of " 0 "；With i-th
A Grad location information PosiWith j-th of Grad location information PosjAs one group of Grad directional information, so as to obtain 6
Group Grad directional information, wherein any one group of Grad directional information is denoted as Orit, 1≤t≤6；
Coarse search module is according to t group Grad directional informations OritIn Grad corresponding to non-" 0 " label it is right respectively
Original block RU is split, so as to obtain KtA coarse search divides block, wherein, k-th of coarse search segmentation block is by original block RU
The two wedge blocks composition obtained after segmentation, in the present embodiment, believes according to 6 groups of Grad directions after gradient pre-filtering
Breath, only remaining 1 coarse search segmentation block, as shown in figure 5, being coarse search segmentation block；Block is divided according to k-th of coarse search,
Mean value computation is carried out respectively according to two wedge blocks to the pixel value in original block RU, obtains mean value and the filling of two wedge blocks
Into corresponding wedge block, so as to form k-th of coarse search prediction block, in the present embodiment, as shown in fig. 6, as shown in Fig. 5
Coarse search segmentation block corresponding to coarse search prediction block；The coarse search rate distortion costs value of k-th of coarse search prediction block is calculated,
So as to obtain KtThe coarse search rate distortion costs value of a coarse search prediction block, in the present embodiment, as shown in fig. 7, in as Fig. 6
The coarse search rate distortion costs value that coarse search prediction block is calculated, value 19；It is synchronous to calculate 6 groups of Grad directional information institutes
Corresponding coarse search rate distortion costs value, and therefrom select the coarse search prediction corresponding to minimum coarse search rate distortion costs value
Block is as optimal coarse search prediction block, in the present embodiment, it is seen that minimum coarse search rate distortion costs value is 19, therefore its is corresponding
Coarse search prediction block is most has coarse search prediction block, i ≠ j, 1≤j≤4；
Two Grad of the smart search module according to optimal coarse search prediction block on split position, find adjacent respectively
Two Grad, so as to according to a Grad and its adjacent two Grad and another Grad and its adjacent two
Grad is respectively split optimal coarse search prediction block, 8 essence search segmentation blocks is obtained, in the present embodiment, such as Fig. 8 institutes
Show, the optimal coarse search prediction block as obtained in coarse search module and its 8 adjacent essence search segmentation blocks；Wherein,
Two wedge blocks that s-th of essence search segmentation block obtains after being divided by original block RU form；According to s-th of essence search segmentation
Block carries out mean value computation according to two wedge blocks to the pixel value in original block RU, obtains the mean value of two wedge blocks simultaneously respectively
It is filled into corresponding wedge block, so as to form s-th of essence search prediction block, in the present embodiment, as shown in figure 9, as root
8 essence search prediction blocks being calculated according to 8 essence search segmentation blocks；The smart searching rate for calculating s-th of essence search prediction block is lost
True cost value, so as to obtain the essence search rate distortion costs value of 8 essence search prediction blocks, in the present embodiment, this 8 essence search
Rate distortion costs value is respectively 82,141,368,421,235,461,159,236, from this 8 essence search rate distortion costs values and
Select the prediction block corresponding to minimum rate distortion costs value pre- as optimal essence search in minimum coarse search rate distortion costs value
Block is surveyed, in the present embodiment, minimum in this 9 rate distortion costs values be still minimum coarse search rate distortion costs value is 19,
I.e. optimal essence search prediction block is identical with optimal coarse search prediction block, finally searches for prediction block according to optimal essence and is counted with original block RU
Calculation obtains residual block, such as Fig. 9, so as to search for the data compression biography of prediction block and residual block realization to original block RU with optimal essence
It is defeated；1≤s≤8.
In the present embodiment, in a kind of 3D coding and decoding videos based on boundary gradient the coding method of depth modelling pattern be by
Following steps carry out：
Step 1,4N × 4N regions in any depth image of note pixel value be original block RU, wherein N is positive integer, 1
≤ N≤8, in the present embodiment, as shown in Figure 1, the pixel value of the original block RU in as one 4 × 4 regions；Calculate original block RU
The borderline 4N-1 Grad in four, upper and lower, left and right, so that one 4 × (4N-1) a gradients are obtained on four boundaries
Value；
All Grad are all added by step 2, obtain Grad and Sum, then by Grad and the binary form of Sum
(3+N) position is moved to right up to mode, obtains gradient pre-filtering threshold value Thr；
Whether four borderline Grad are judged all less than gradient pre-filtering threshold value Thr, if so, not to four sides
Grad in boundary makes any change, and otherwise, the Grad that gradient pre-filtering threshold value Thr is less than on four boundaries is all put
" 0 ", without modification, in the present embodiment, one shares 12 ladders to the Grad more than gradient pre-filtering threshold value Thr on four boundaries
It is all added, obtains Grad and Sum=2+1+45+1+0+1+44+0+0+2+1+1=98 by angle value, then by Grad and
The binary expression mode 1100010 of Sum moves to right (3+1)=4, obtains the binary expression mode of gradient pre-filtering threshold value
Thr=110, i.e. Thr=6, it is clear that four borderline Grad are not all of being less than gradient pre-filtering threshold value Thr, therefore by four
On a boundary be less than 6 Grad whole reset, the Grad more than 6 without modification, as shown in figure 3, four are borderline non-
The 44 of 45 and lower boundary of the only remaining coboundary of " 0 " Grad；
Step 3, to i-th, borderline 4N-1 Grad be successively into line label, and borderline 4N-1 to i-th
Grad carries out descending sort, obtains the corresponding sequence that Grad corresponds to label, is denoted as i-th of Grad location information Posi,
In the present embodiment, as shown in figure 4, as 4 borderline Grad location informations of original block, wherein, Grad is " 0 "
Label be also denoted as " 0 ", 1≤i≤4；
Step 4, with i-th of Grad location information PosiWith j-th of Grad location information PosjAs one group of Grad
Directional information, so as to obtain 6 groups of Grad directional informations, wherein any one group of Grad directional information is denoted as Orit, 1≤t≤
6；I ≠ j, 1≤j≤4；
Step 5, initialization t=1；
Step 6, according to t group Grad directional informations OritIn Grad corresponding to non-" 0 " label respectively to original
Block RU is split, so as to obtain KtA coarse search divides block, in the present embodiment, as shown in figure 5, being coarse search segmentation
Block, wherein, two wedge blocks that k-th of coarse search segmentation block obtains after being divided by original block RU form；
Step 7 divides block according to k-th coarse search, to the pixel value in original block RU according to two wedge blocks respectively into
Row mean value computation obtains the mean value of two wedge blocks and is filled into corresponding wedge block, so as to form k-th of coarse search prediction
Block, as shown in fig. 6, being the coarse search prediction block obtained in the present embodiment；
Step 8, the coarse search rate distortion costs value for calculating k-th of coarse search prediction block, so as to obtain KtA coarse search is pre-
The coarse search rate distortion costs value of block is surveyed, in the present embodiment, as shown in fig. 7, the coarse search rate distortion of the coarse search prediction block
Cost value is 19；
T+1 is assigned to t by step 9, judges whether t ＞ 6 are true, if so, then perform step 9；Otherwise return to step 5；
Step 10 is selected from all coarse search rate distortion costs values corresponding to minimum coarse search rate distortion costs value
Coarse search prediction block is as optimal coarse search prediction block, in the present embodiment, it is clear that minimum coarse search rate distortion costs value is
19, therefore the coarse search prediction block corresponding to 19 is optimal coarse search prediction block；
Step 11, two Grad according to optimal coarse search prediction block on split position, find adjacent two respectively
A Grad；
Step 12, according to a Grad and its adjacent two Grad and another Grad and its adjacent two
Grad is respectively split optimal coarse search prediction block, 8 essence search segmentation blocks is obtained, in the present embodiment, such as Fig. 8 institutes
Show, as this 8 essence search segmentation blocks；Wherein, s-th of essence search segmentation block is two wedges obtained after being divided by original block RU
Shape block forms；
Step 13 divides block according to s-th coarse search, to the pixel value in original block RU according to two wedge blocks respectively into
Row mean value computation obtains the mean value of two wedge blocks and is filled into corresponding wedge block, so as to form s-th of essence search prediction
Block, in the present embodiment, as shown in figure 9, being the 8 essence search prediction blocks obtained according to 8 essence search segmentation blocks；
Step 14, the essence search rate distortion costs value for calculating s-th of essence search prediction block, it is pre- so as to obtain 8 essence search
The essence search rate distortion costs value of block is surveyed, from this 8 essence search rate distortion costs values and minimum coarse search rate distortion costs value
The prediction block corresponding to minimum rate distortion costs value is selected as optimal essence search prediction block, in the present embodiment, this 8 essences
It is respectively 82,141,368,421,235,461,159,236 to search for rate distortion costs value, and minimum coarse search rate distortion costs value is
19, therefore optimal essence search prediction block is the optimal coarse search prediction block corresponding to minimum coarse search rate distortion costs value；
Residual block is calculated according to optimal essence search prediction block and original block RU in step 15, in the present embodiment, such as schemes
Shown in 10, the as calculating process of residual block, so as to search for prediction block with optimal essence and residual block realizes number to original block RU
According to compression transmission；1≤s≤8.
The present invention carries out the circuit design based on FPGA of intra-frame predictive encoding method for the original block of 4 × 4 sizes, adopts
Behavioral scaling description is carried out with Verilog HDL, is imitated based on Xilinx XC6VLX760 FPGA development boards using ISE softwares
True and comprehensive, the present invention is tested using depth image block as shown in Figure 1, is existed compared to Gustavo Sanchez et al.
" the Low- delivered in 201730th Symposium on Integrated Circuits and Systems Design meetings
Area Scalable Hardware Architecture for DMM-1 Encoder of 3D-HEVC Video Coding
Circuit in Standard ", for the original block of 4 × 4 sizes, predetermined period number is 132 periods, and the present invention only needs
In 16 periods, it is equivalent to its 12.1%.
Claims (2)
1. the coding circuit of depth modelling pattern, remembers in any depth image in a kind of 3D coding and decoding videos based on boundary gradient
4N × 4N regions pixel value for original block RU, wherein N is positive integer, 1≤N≤8；It is characterized in that：The coding circuit packet
It includes：Input data processing module, coarse search module, smart search module and wedge block model data store module；
The input data processing module receives externally input original block RU, and calculate the original block RU it is upper and lower, left,
Right four borderline 4N-1 Grad, so as to which one is obtained 4 × (4N-1) a Grad on four boundaries, by all ladders
Angle value is all added, and obtains Grad and Sum, then the binary expression mode of the Grad and Sum is moved to right (3+N) position,
Obtain gradient pre-filtering threshold value Thr；Judge to whether there is the gradient less than the gradient pre-filtering threshold value Thr on four boundaries
Value, if in the presence of the Grad whole reset of the gradient pre-filtering threshold value Thr will be less than on four boundaries；Again to i-th of side
4N-1 Grad in boundary is successively into line label；Described i-th borderline 4N-1 Grad is subjected to descending sort, from
And the corresponding sequence that Grad corresponds to label is obtained, it is denoted as i-th of Grad location information Posi, wherein, Grad is " 0 "
Label is also denoted as " 0 ", 1≤i≤4；With i-th of Grad location information PosiWith j-th of Grad location information PosjAs one
Group Grad directional information, so as to obtain 6 groups of Grad directional informations, wherein any one group of Grad directional information is denoted as
Orit, 1≤t≤6；
The coarse search module is according to t group Grad directional informations OritIn Grad corresponding to non-" 0 " label respectively to institute
It states original block RU to be split, so as to obtain KtA coarse search divides block, wherein, k-th of coarse search segmentation block is by the original
The two wedge blocks composition obtained after beginning block RU segmentations；Block is divided according to k-th of coarse search, to the pixel in the original block RU
Value carries out mean value computation respectively according to two wedge blocks, obtains the mean value of two wedge blocks and is filled into corresponding wedge block,
So as to form k-th of coarse search prediction block；The coarse search rate distortion costs value of k-th of coarse search prediction block is calculated, so as to obtain
KtThe coarse search rate distortion costs value of a coarse search prediction block；The synchronous coarse search calculated corresponding to 6 groups of Grad directional informations
Rate distortion costs value, and the coarse search prediction block corresponding to minimum coarse search rate distortion costs value is therefrom selected as optimal thick
Search for prediction block, i ≠ j, 1≤j≤4；
Two Grad of the essence search module according to optimal coarse search prediction block on split position, find adjacent respectively
Two Grad, so as to according to a Grad and its adjacent two Grad and another Grad and its adjacent two
Grad is respectively split the optimal coarse search prediction block, obtains 8 essence search segmentation blocks；Wherein, s-th of essence search
Two wedge blocks that segmentation block obtains after being divided by the original block RU form；According to s-th of essence search segmentation block, to described
Pixel value in original block RU carries out mean value computation respectively according to two wedge blocks, obtains the mean value of two wedge blocks and is filled into
In corresponding wedge block, so as to form s-th of essence search prediction block；Calculate the smart searching rate distortion of s-th of essence search prediction block
Cost value, so as to obtain the essence search rate distortion costs value of 8 essence search prediction blocks, from this 8 essence search rate distortion costs values
With the prediction block corresponding to minimum rate distortion costs value is selected in minimum coarse search rate distortion costs value as optimal essence search
Finally residual block is calculated according to optimal essence search prediction block and the original block RU in prediction block, so as to it is described most
Excellent essence search prediction block and residual block realize that the data compression to the original block RU is transmitted；1≤s≤8.
2. a kind of coding method of depth modelling pattern in 3D coding and decoding videos based on boundary gradient, it is characterized in that by following step
It is rapid to carry out：
Step 1,4N × 4N regions in any depth image of note pixel value be original block RU, wherein N is positive integer, 1≤N≤
8；The borderline 4N-1 Grad in four, upper and lower, left and right of the original block RU is calculated, so as to be had altogether on four boundaries
Obtain 4 × (4N-1) a Grad；
All Grad are all added by step 2, obtain Grad and Sum, then by the Grad and the binary form of Sum
(3+N) position is moved to right up to mode, obtains gradient pre-filtering threshold value Thr；
Judge to whether there is the Grad less than gradient pre-filtering threshold value Thr on four boundaries, if in the presence of will be on four boundaries
Less than the Grad whole reset of gradient pre-filtering threshold value Thr；
Step 3,4N-1 Grad borderline to described i-th are successively into line label, and borderline 4N-1 is a to i-th
Grad carries out descending sort, obtains the corresponding sequence that Grad corresponds to label, is denoted as i-th of Grad location information Posi,
Wherein, Grad is also denoted as " 0 ", 1≤i≤4 for the label of " 0 "；
Step 4, with i-th of Grad location information PosiWith j-th of Grad location information PosjAs one group of Grad direction
Information, so as to obtain 6 groups of Grad directional informations, wherein any one group of Grad directional information is denoted as Orit, 1≤t≤6；i≠
J, 1≤j≤4；
Step 5, initialization t=1；
Step 6, according to t group Grad directional informations OritIn Grad corresponding to non-" 0 " label respectively to the original block
RU is split, so as to obtain KtA coarse search divides block, wherein, k-th of coarse search segmentation block is by the original block RU points
The two wedge blocks composition obtained after cutting；
Step 7 divides block according to k-th coarse search, to the pixel value in the original block RU according to two wedge blocks respectively into
Row mean value computation obtains the mean value of two wedge blocks and is filled into corresponding wedge block, so as to form k-th of coarse search prediction
Block；
Step 8, the coarse search rate distortion costs value for calculating k-th of coarse search prediction block, so as to obtain KtA coarse search prediction block
Coarse search rate distortion costs value；
T+1 is assigned to t by step 9, judges whether t ＞ 6 are true, if so, then perform step 9；Otherwise return to step 5；
Step 10 selects slightly searching corresponding to minimum coarse search rate distortion costs value from all coarse search rate distortion costs values
Rope prediction block is as optimal coarse search prediction block；
Step 11, two Grad according to optimal coarse search prediction block on split position, find two adjacent ladders respectively
Angle value；
Step 12, according to a Grad and its adjacent two Grad and another Grad and its adjacent two gradients
Value is respectively split the optimal coarse search prediction block, obtains 8 essence search segmentation blocks；Wherein, s-th of essence search segmentation
Two wedge blocks that block obtains after being divided by the original block RU form；
Step 13 divides block according to s-th coarse search, to the pixel value in the original block RU according to two wedge blocks respectively into
Row mean value computation obtains the mean value of two wedge blocks and is filled into corresponding wedge block, so as to form s-th of essence search prediction
Block；
Step 14, the essence search rate distortion costs value for calculating s-th of essence search prediction block, so as to obtain 8 essence search prediction blocks
Essence search rate distortion costs value, selected from this 8 essence search rate distortion costs values and minimum coarse search rate distortion costs value
Prediction block corresponding to minimum rate distortion costs value is as optimal essence search prediction block；
Residual block is calculated according to optimal essence search prediction block and the original block RU in step 15, so as to it is described most
Excellent essence search prediction block and residual block realize that the data compression to the original block RU is transmitted；1≤s≤8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810060934.2A CN108271025B (en) | 2018-01-22 | 2018-01-22 | The coding circuit of depth modelling mode and its coding method in 3D coding and decoding video based on boundary gradient |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810060934.2A CN108271025B (en) | 2018-01-22 | 2018-01-22 | The coding circuit of depth modelling mode and its coding method in 3D coding and decoding video based on boundary gradient |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108271025A true CN108271025A (en) | 2018-07-10 |
| CN108271025B CN108271025B (en) | 2019-11-29 |
Family
ID=62776391
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810060934.2A Active CN108271025B (en) | 2018-01-22 | 2018-01-22 | The coding circuit of depth modelling mode and its coding method in 3D coding and decoding video based on boundary gradient |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108271025B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103997645A (en) * | 2014-04-29 | 2014-08-20 | 长沙理工大学 | Quick HEVC intra-frame coding unit and pattern decision method |
| US20150195540A1 (en) * | 2014-01-03 | 2015-07-09 | Futurewei Technologies, Inc. | System and Method for Estimating View Synthesis Distortion |
| CN105120290A (en) * | 2015-08-04 | 2015-12-02 | 宁波大学 | Fast coding method for depth video |
| CN105791862A (en) * | 2016-03-21 | 2016-07-20 | 杭州电子科技大学 | Three-dimensional video coding depth map internal mode selection method based on edge complexity |
| CN106331727A (en) * | 2016-08-26 | 2017-01-11 | 天津大学 | Simplified search method for depth modeling modes |
-
2018
- 2018-01-22 CN CN201810060934.2A patent/CN108271025B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150195540A1 (en) * | 2014-01-03 | 2015-07-09 | Futurewei Technologies, Inc. | System and Method for Estimating View Synthesis Distortion |
| CN103997645A (en) * | 2014-04-29 | 2014-08-20 | 长沙理工大学 | Quick HEVC intra-frame coding unit and pattern decision method |
| CN105120290A (en) * | 2015-08-04 | 2015-12-02 | 宁波大学 | Fast coding method for depth video |
| CN105791862A (en) * | 2016-03-21 | 2016-07-20 | 杭州电子科技大学 | Three-dimensional video coding depth map internal mode selection method based on edge complexity |
| CN106331727A (en) * | 2016-08-26 | 2017-01-11 | 天津大学 | Simplified search method for depth modeling modes |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108271025B (en) | 2019-11-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103260031B (en) | Method and apparatus for encoding/decoding to motion vector | |
| CN104994380B (en) | Method and apparatus for being coded and decoded to motion vector | |
| CN104735452B (en) | By the method and apparatus using big converter unit image to be encoded and decodes | |
| CN103220521B (en) | The coding unit of image boundary is carried out method and the equipment of Code And Decode | |
| CN104796694B (en) | Optimization intraframe video coding method based on video texture information | |
| CN102497546B (en) | Screen image compression device and compression method thereof | |
| CN104038760B (en) | A kind of wedge shape Fractionation regimen system of selection of 3D video depths image frame in and system | |
| CN108737841A (en) | Coding unit depth determination method and device | |
| US9883200B2 (en) | Method of acquiring neighboring disparity vectors for multi-texture and multi-depth video | |
| CN105898332B (en) | For the fast deep figure frame mode decision method of 3D-HEVC coding standards | |
| CN107623848B (en) | A kind of method for video coding and device | |
| CN104702955A (en) | Method and system for predictive coding of screen video | |
| CN104754359A (en) | Depth map coding distortion forecasting method for two-dimensional free viewpoint video | |
| CN110677644B (en) | Video coding and decoding method and video coding intra-frame predictor | |
| CN105263026B (en) | Global vector acquisition methods based on probability statistics and image gradient information | |
| CN101835047A (en) | Fast UMHexagonS motion estimation algorithm based on residual error reduction rate | |
| CN108924551A (en) | The prediction technique and relevant device of video image coding pattern | |
| CN108271025B (en) | The coding circuit of depth modelling mode and its coding method in 3D coding and decoding video based on boundary gradient | |
| CN105007494B (en) | Wedge-shaped Fractionation regimen selection method in a kind of frame of 3D video depths image | |
| Wang et al. | HEVC intra frame based compressed domain video summarization | |
| CN108134940B (en) | Coding circuit and method applied to depth image intra prediction in Video coding | |
| CN110049339A (en) | Prediction direction choosing method, device and storage medium in image coding | |
| CN108449591A (en) | A kind of deep video DMM patterns simplification method | |
| CN103997635B (en) | The synthesis viewpoint distortion prediction method of free viewpoint video and coding method | |
| CN108012151B (en) | A kind of adaptive motion vector precision fast selecting method and device based on Bayes's classification |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |