CN107197253A - A kind of quick decision methods of QTBT based on KB wave filters and system - Google Patents
A kind of quick decision methods of QTBT based on KB wave filters and system Download PDFInfo
- Publication number
- CN107197253A CN107197253A CN201710229733.6A CN201710229733A CN107197253A CN 107197253 A CN107197253 A CN 107197253A CN 201710229733 A CN201710229733 A CN 201710229733A CN 107197253 A CN107197253 A CN 107197253A
- Authority
- CN
- China
- Prior art keywords
- image
- mtd
- high frequency
- frequency points
- mrow
- 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
- 238000000034 method Methods 0.000 title claims abstract description 45
- 238000001914 filtration Methods 0.000 claims abstract description 37
- 238000012545 processing Methods 0.000 claims abstract description 8
- 230000009191 jumping Effects 0.000 claims abstract description 5
- 102000008297 Nuclear Matrix-Associated Proteins Human genes 0.000 claims description 16
- 108010035916 Nuclear Matrix-Associated Proteins Proteins 0.000 claims description 16
- 210000000299 nuclear matrix Anatomy 0.000 claims description 16
- 238000005192 partition Methods 0.000 description 14
- 238000013461 design Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 238000000205 computational method Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 210000001519 tissue Anatomy 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
- H04N19/117—Filters, e.g. for pre-processing or post-processing
-
- 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
- H04N19/10—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
-
- 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
- H04N19/189—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the adaptation method, adaptation tool or adaptation type used for the adaptive coding
- H04N19/192—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the adaptation method, adaptation tool or adaptation type used for the adaptive coding the adaptation method, adaptation tool or adaptation type being iterative or recursive
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 quick decision methods of the QTBT based on KB wave filters and system, the system includes acquisition module, chooses module, searching modul and judge module.This method includes obtaining filtering image after being filtered video frame images processing using KB wave filters；Choose corresponding high frequency points threshold value；Find out the high frequency points maximum of image-region；Judge whether the high frequency points maximum found out is less than the high frequency points threshold value selected, if so, then jumping out current coded unit divides recursive program；Conversely, then proceeding to divide recurrence to current coded unit according to QTBT algorithms.By using the method and system of the present invention, whether CU can be proceeded to divide recurrence progress anticipation based on characteristics of image, greatly reduce video encoding time.The present invention can be widely applied in QTBT algorithm field of video encoding as the quick decision methods of QTBT and system based on KB wave filters.
Description
Technical field
The present invention relates to video coding technique, more particularly to a kind of quick decision methods of QTBT based on KB wave filters and it is
System, it is applied in video encoding unit CU partition process.
Background technology
Technology word is explained：
KB wave filters：Full name is Ker-Bohme wave filters, and it is a high-pass filter, because it is extracting characteristics of image
Aspect has superior function, and it is used widely in image latent writing analysis.
, the Video Coding Experts Group (VCEG) and world mark of international telecommunication union telecommunication's Standardization Sector (ITU-T) in 2015
The motion picture expert group (MPEG) of standardization tissue/International Electrotechnical Commission (ISO/IEC) has set up Joint Video
Exploration Team (JVET) joint video research work group, the working group are primarily to coding and decoding video of future generation is done
Prepare.The reference software JEM of the working group is a series of improvement done on HEVC reference software HM, on original HM basis
On, JVET working groups with the addition of many new instruments wherein, for example：QTBT (Quadtree plus binary tree) is calculated
Method, OBMC (Overlapped block motion compensation) algorithm etc..In addition, with high-resolution video
Popularization, original 64x64, which is divided, can not meet the division demand of HD video, so for Video coding mark of future generation
, it is necessary to support bigger code tree unit CTU (coding tree unit) to divide for standard, therefore in JEM4.0, acquiescence
Code tree unit CTU be 128x128, while can also support bigger 256x256 types.In addition, intra prediction mode is also carried
It is high to 67 kinds of predictive modes and to add the filtering interpolations of four taps, it can preferably improve precision of prediction.As this is
The addition of row instrument, JEM4.0 is compared with HM, and highest can obtain 28% BD-Rate (RA (random in performance
Access) random access configuration) performance increase, but while performance increase is brought, the complexity of encoder is also with carrying
Height, thus efficiently reduce the scramble time be also the problem of must pay attention to during preparing video encoding standard of future generation it
One.
In JEM, compared with HM, change maximum is exactly QTBT algorithms, and it changes the division of original coding and decoding video
Coding unit CU (coding unit) overall framework, the combination of binary tree and quaternary tree is changed into by original quad-tree partition,
As shown in figure 1, a code tree unit CTU can be divided into four coding unit CU by square, also can be by rectangular partition into 2
Coding unit CU, is divided into vertical division and horizontal division, and coding unit CU once uses rectangular partition, then no longer carries out square
Quad-tree partition, this dividing mode compare before quad-tree partition for it is more flexibly and more fine, can give
Encoder brings average 4% BD-Rate coding efficiencies lifting, but simultaneously as there is the trial of more kinds of divisions, encodes
Time is also increase at double, so having the high performance coding tools of high complexity for this, the design, which is proposed, to be based on
The quick decision algorithms of QTBT of Ker-Bohme wave filters reduce the effect of scramble time to reach.
The content of the invention
In order to solve the above-mentioned technical problem, quickly adjudicated it is an object of the invention to provide a kind of QTBT based on KB wave filters
Method, so that coding unit CU division efficiency is improved, to reach the effect for reducing the scramble time.
It is an object of the invention to provide a kind of quick decision systems of the QTBT based on KB wave filters, so as to improve coding unit
CU division efficiency, to reach the effect for reducing the scramble time.
The technical solution adopted in the present invention is：A kind of quick decision methods of QTBT based on KB wave filters, this method
Step includes：
The first image is obtained, wherein, described first image refers to entering video frame images using Ker-Bohme wave filters
The filtering image obtained after row filtering process；
According to the block size of current coded unit, corresponding high frequency points threshold value is chosen；
The high frequency points maximum of image-region is found out, wherein, described image region is referred in the first image with working as
The corresponding image-region of preceding coding unit；
Judge whether the high frequency points maximum found out is less than the high frequency points threshold value selected, if so, then jumping out current volume
Code dividing elements recursive program；Conversely, then proceeding to divide recurrence to current coded unit according to QTBT algorithms.
Further, the step for the first image of the acquisition, it is specifically included：
Obtain expanding image after video frame images to be expanded to 1 row and 1 row in a mirror-image fashion；
The nuclear matrix of Ker-Bohme wave filters obtained into filtering image with expanding after image carries out convolution, it is described to obtain
The first image that filtering image is obtained for needed for.
Further, the nuclear matrix KB of the Ker-Bohme wave filters is as follows：
Further, the step of this method also includes the step for high frequency points threshold value is obtained, and the high frequency points threshold value is obtained
The step for include：
Multiple filtering samples are obtained after being filtered out respectively to multiple Sample video two field pictures using Ker-Bohme wave filters
This image；
According to default block size, cutting division is carried out to multiple filtered samples images, so as to obtain multiple rectangular images
Block；
The lookup of high frequency points maximum is carried out to each rectangular image block, so as to find out in each rectangular image block
High frequency points maximum；
According to the numerical value of the multiple high frequency points maximums found out, so as to be ranked up to multiple rectangular image blocks；
Query search is carried out to the rectangular image block after sequence by the way of binary chop, met so as to find out
The rectangular image block of preparatory condition；
It regard the high frequency points maximum in the rectangular image block found out as the high frequency points corresponding with default block size
Threshold value.
Further, the default block size include 8x8,8x16,16x8,8x32,32x8,16x16,16x32,
32x16,32x32,16x64,64x16,32x64,64x32 and 64x64.
Another technical scheme of the present invention is：A kind of quick decision systems of QTBT based on KB wave filters, this is
System includes：
Acquisition module, for obtaining the first image, wherein, described first image refers to using Ker-Bohme wave filters
Video frame images are filtered with the filtering image obtained after processing；
Module is chosen, for the block size according to current coded unit, corresponding high frequency points threshold value is chosen；
Searching modul, the high frequency points maximum for finding out image-region, wherein, described image region is referred to
The image-region corresponding with current coded unit in one image；
Whether judge module, the high frequency points maximum for judging to find out is less than the high frequency points threshold value selected, if so,
Then jump out current coded unit and divide recursive program；Conversely, then proceeding to divide to current coded unit according to QTBT algorithms
Recurrence.
Further, the acquisition module is specifically included：
Image expands submodule, for obtaining expanding image after video frame images to be expanded to 1 row and 1 row in a mirror-image fashion；
Submodule is filtered, for the nuclear matrix of Ker-Bohme wave filters to be filtered with expanding after image carries out convolution
Image, the first image that the obtained filtering image is obtained for needed for.
Further, the nuclear matrix KB of the Ker-Bohme wave filters is as follows：
Further, the system also includes high frequency points threshold value acquisition module, and the high frequency points threshold value acquisition module is specifically included：
First acquisition submodule, for being filtered respectively to multiple Sample video two field pictures using Ker-Bohme wave filters
Ripple obtains multiple filtered samples images after going out；
Second acquisition submodule, for according to default block size, cutting division to be carried out to multiple filtered samples images, from
And obtain multiple rectangular image blocks；
3rd acquisition submodule, the lookup for carrying out high frequency points maximum to each rectangular image block, so as to search
The high frequency points maximum gone out in each rectangular image block；
Sorting sub-module, for the numerical value according to the multiple high frequency points maximums found out, so as to multiple rectangular images
Block is ranked up；
Submodule is searched, for being searched by the way of binary chop to carry out inquiry to the rectangular image block after sequence
Rope, so as to find out the rectangular image block for meeting preparatory condition；
Threshold value determination sub-module, for the high frequency points maximum in the rectangular image block that will find out as with default piece
The corresponding high frequency points threshold value of size.
Further, the default block size include 8x8,8x16,16x8,8x32,32x8,16x16,16x32,
32x16,32x32,16x64,64x16,32x64,64x32 and 64x64.
The beneficial effects of the invention are as follows：By using method of the invention, it is possible in the coding unit of progress QTBT algorithms
Divide in recursive procedure, whether need to carry out further recurrence to coding unit and divide to carry out anticipation, when judging present encoding
When unit need not proceed to divide recurrence, then jump out current coded unit and divide recursive program, conversely, then according to QTBT algorithms
Current coded unit is proceeded to divide recurrence, it can be seen that, in the coding unit partition process of QTBT algorithms, by making
With the inventive method, it can reduce coding unit on the basis of guarantee does not reduce video encoding quality and divide recursive operation,
The time of Video coding is reduced, so as to greatly improve the efficiency of Video coding.
The present invention another beneficial effect be：Using the system of the present invention recurrence mistake is divided in the coding unit of QTBT algorithms
Cheng Zhong, whether to coding unit need carry out further recurrence divide carry out anticipation, when judging present encoding list if can realize
When member need not proceed to divide recurrence, then jump out current coded unit and divide recursive program, conversely, then according to QTBT algorithms pair
Current coded unit proceed divide recurrence, it can be seen that, by using present system QTBT algorithms coding unit
In partition process, it can reduce coding unit on the basis of guarantee does not reduce video encoding quality and divide recursive operation, reduce
The time of Video coding, so as to greatly improve the efficiency of Video coding.
Brief description of the drawings
Fig. 1 is the division recursive principle schematic diagram of QTBT algorithms；
Fig. 2 is a kind of step schematic flow sheet of the quick decision methods of QTBT based on KB wave filters of the present invention；
Fig. 3 is a kind of structured flowchart of the quick decision systems of QTBT based on KB wave filters of the present invention；
Fig. 4 is an a kind of specific embodiment steps flow chart of the quick decision methods of QTBT based on KB wave filters of the present invention
Figure；
Fig. 5 is the principle schematic that using KB wave filters video frame images are filtered with processing；
Fig. 6 is the artwork of video frame images and the contrast schematic diagram of filtering image；
Fig. 7 is the schematic diagram of cloud high frequency points in filtering image；
Fig. 8 is the principle schematic in the coding unit of QTBT algorithms divides recursive procedure using the present invention.
Embodiment
As shown in Fig. 2 the step of a kind of quick decision methods of QTBT based on KB wave filters, this method includes：
The first image is obtained, wherein, described first image refers to entering video frame images using Ker-Bohme wave filters
The filtering image obtained after row filtering process；
According to the block size of current coded unit, corresponding high frequency points threshold value is chosen；
The high frequency points maximum of image-region is found out, wherein, described image region is referred in the first image with working as
The corresponding image-region of preceding coding unit；
Judge whether the high frequency points maximum found out is less than the high frequency points threshold value selected, if so, then jumping out current volume
Code dividing elements recursive program；Conversely, then proceeding to divide recurrence to current coded unit according to QTBT algorithms.
As shown in figure 3, a kind of quick decision systems of QTBT based on KB wave filters, the system includes：
Acquisition module, for obtaining the first image, wherein, described first image refers to using Ker-Bohme wave filters
Video frame images are filtered with the filtering image obtained after processing；
Module is chosen, for the block size according to current coded unit, corresponding high frequency points threshold value is chosen；
Searching modul, the high frequency points maximum for finding out image-region, wherein, described image region is referred to
The image-region corresponding with current coded unit in one image；
Whether judge module, the high frequency points maximum for judging to find out is less than the high frequency points threshold value selected, if so,
Then jump out current coded unit and divide recursive program；Conversely, then proceeding to divide to current coded unit according to QTBT algorithms
Recurrence.
Further it is specifically described with reference to specific examples below to be done to the present invention.
For coding and decoding video, general flat place, the coding unit CU of division can be bigger, conversely, right
Division at details, then can be divided into multiple lower Item unit CU, that is, can divide it is more fine, to ensure to regard
Frequency information can be completely transmitted.In JEM, its incessantly only have quad-tree partition, can also exist binary tree divide, it is necessary to
The division of trial becomes many, that is, needs the rate distortion costs RD-Cost done checking process also to accordingly increase.Therefore, it is based on
The QTBT that characteristics of image carrys out decision-making coding unit CU in advance is divided, and can be carried on the basis of guarantee does not reduce video encoding quality
It is preceding to jump out division recurrence without the current coded unit CU divided again.For this principle, it is proposed that the present invention program.
The design principle of the present invention is on the premise of based on Ker-Bohme filtering current video frames, according to obtained figure
As feature, whether high-speed decision needs further to divide coding unit CU recurrence, if reaching the design requirement, jumps out current
Coding unit divides recursive program, to reach the purpose for reducing the scramble time.
As shown in figure 4, the step of a kind of quick decision methods of QTBT based on KB wave filters, this method institute specific as follows
Show.
Part I：Ker-Bohme filtering process
S101, the first image of acquisition, wherein, described first image is referred to using Ker-Bohme wave filters to frame of video
Image is filtered the filtering image obtained after processing, and the filtering image is a pixel map.
Present invention design is the method quickly adjudicated based on characteristics of image is extracted, therefore, for characteristics of image
Extract, realized in the present invention using a kind of effective high-pass filter-Ker-Bohme wave filters, wherein, the Ker-
The nuclear matrix KB of Bohme wave filters is as follows：
Then, can by using this nuclear matrix KB and the filtering image obtained after original video frame images progress convolution
Effectively to retain the high-frequency signal of image, wherein, the formula of the convolution is：Z=KB*Y；
Therefore, for step S101, it implements step and included：
S1011, video frame images are expanded to 1 row and 1 row in a mirror-image fashion after obtain expanding image；
S1012, will the nuclear matrix of Ker-Bohme wave filters with expand image carry out convolution after obtain filtering image, it is described
The first image that obtained filtering image is obtained for needed for；
As shown in figure 5, what is represented in solid box is original video frame images picture, what is represented in dotted line frame is to use KB cores
Matrix carries out the process of convolution to image pixel, and KB nuclear matrix will travel through whole according to the direction of arrow and expand picture progress convolution,
To obtain and the correspondingly sized filtering image of original block (original video frame images) size；
For example, in Figure 5 KB nuclear matrix and top left corner pixel do convolution obtain filtering image first pixel p (0,
0) process, is implemented as shown in following first formula：
Then, KB cores can move right by step-length for 1 distance, and with expanding the pixel progress convolution of image, press
According to the method for the first formula, the filtered pixel value of the first row is obtained, until the 3rd row of KB nuclear matrix are with expanding the last of image
One row are overlapped, then KB nuclear matrix, will move down 1 row by step-length for 1 distance, and return to the first row for expanding pixel, weight
Newly according to the mode that moves right before, the filtered pixel value of the second row is obtained, in this way, will be 1 by step-length after KB nuclear matrix
Distance move down to the right, respectively with expand image pixel carry out convolution, according to the computational methods of above-mentioned first formula,
It finally just can obtain one and artwork filtering image of a size；(left side is artwork, and the right is schemed for filtering as shown in Figure 6
Picture), after filtering after, can substantially observe the profile of image, and retain the cloud in the upper left corner in the details of image, such as Fig. 6
Piece, in image after the filtering, high frequency points therein are still can see, as shown in Figure 7.As can be seen here, obtain filtered
After image pixel, that is, obtain after the first image, just can obtain specific high-frequency information, so can when selection block is divided
To make full use of these high-frequency informations to be used for anticipation, so as to quickly judge whether coding unit CU can need not do further recurrence
Divide, reach on the basis of guarantee does not reduce video encoding quality, reduce the effect of scramble time.
Part II：The quick decision methods of QTBT
According to the content of Part I, codec first carries out mirror image expansion, so before each frame is encoded to image
KB core filtering is carried out to it afterwards, a filtered pixel map is obtained, and the coding list of coded image can be entered after filtering
First CU divides the stage, starts QTBT blocks recurrence and divides.
S102, in the coding unit CU division stages, according to current coding unit CU block size, choose corresponding
High frequency points threshold value T, wherein, block size is different, and the high frequency points threshold value corresponding to it is different.
The step for being obtained for high frequency points threshold value S100, it is performing realization before QTBT is quickly adjudicated, and it has
Body includes：
S1001, multiple Sample video two field pictures are filtered out respectively using Ker-Bohme wave filters after obtain multiple
Filtered samples image；
S1002, according to default block size, cutting division is carried out to multiple filtered samples images, so as to obtain multiple squares
Shape image block；
S1003, the lookup that high frequency points maximum is carried out to each rectangular image block, so as to find out each histogram
As the high frequency points maximum in block；
S1004, the numerical value according to the multiple high frequency points maximums found out, so as to arrange multiple rectangular image blocks
Sequence；
S1005, by the way of binary chop come to after sequence rectangular image block carry out query search, so as to search
Go out to meet the rectangular image block of preparatory condition, wherein, the preparatory condition is to be regarded as flat histogram without subdivided
As block；
S1006, using the high frequency points maximum in the rectangular image block found out as corresponding with default block size
High frequency points threshold value；
In the present embodiment, Sample video two field picture uses 10000 512x512 image, first by 10000
512x512 Sample video two field picture carries out KB filtering (i.e. using Ker-Bohme wave filters respectively to 10000 respectively
512x512 Sample video two field picture is filtered processing respectively), obtain multiple filtered samples images；Then, according to default
Obtained multiple filtered samples images are carried out cutting division, obtain several different size of piece, for example, need to cut by block size
When being cut into 32x16 blocks, i.e., now default block size is 32x16, can so cut out 5120000 rectangular image blocks；Then,
Due to having a high frequency points maximum in each rectangular image block, therefore, the height in each rectangular image block is found out
Frequency maximum；And then, these rectangular image blocks are ranked up according to the numerical values recited of its own high frequency points maximum；Adopt again
With the mode of binary chop to carry out query search to the rectangular image block after sequence, find out and do not need subdivided can consider
It is flat rectangular image block, and the high frequency value maximum in this rectangular image block is right as current block type 32x16 institutes
The high frequency points threshold value T answered.
For above-mentioned default block size, its include 8x8,8x16,16x8,8x32,32x8,16x16,16x32,
32x16,32x32,16x64,64x16,32x64,64x32 and 64x64, that is, need for the block size shown in below table 1,
To carry out high frequency points Threshold-training.Table 1 is as follows：
Table 1
It can be seen that, final one meets the high frequency points threshold value obtained corresponding to 14 Seed-ginger sizes together.Due to according to QTBT in JEM4.0
Default parameters understand, when block size be 64x8 and 8x64 when, binary tree divide depth necessarily be more than 3, be unsatisfactory for QTBT
The set condition that can proceed with division in algorithm, that is to say, that coding unit CU (the i.e. 64x8 coding units of the block size
And 8x64 coding units) without being further continued for carrying out division detection, it is both feelings of 64x8 and 8x64 for block size therefore
Condition, without obtaining the high frequency points threshold value corresponding to it.In addition, for the high frequency points threshold value corresponding to different block sizes, it should be abided by
The principle followed is：Block is bigger, and high frequency points threshold value needs smaller；For example, high frequency points threshold value of the block size corresponding to 64x64 need to be less than
Block size is the high frequency points threshold value corresponding to 32x32, and block size need to be less than 64x16 institutes for the high frequency points threshold value corresponding to 64x32
Corresponding high frequency points threshold value, therefore, for step S100, it preferably should also include set-up procedure, and the set-up procedure is, when looking into
When the high frequency points threshold value found out does not meet above-mentioned default principle, then the high frequency points threshold value found out is adjusted.
S103, corresponding with current coded unit CU image-region in the first image is found out, the height corresponding to it
Frequency maximum HFPmax；
The high frequency points maximum HFP that S104, judgement are found outmaxWhether the high frequency points threshold value T selected is less than, if so, then
Jump out current coded unit and divide recursive program；Passed conversely, then proceeding division to current coded unit according to QTBT algorithms
Return, check the rate distortion costs RDcost of each division, until finding optimum division mode, as shown in left in Figure 8.Wherein, institute
The division recurrence stated includes binary tree horizontal division recurrence, binary tree vertical division recurrence and quad-tree partition recurrence.
Obtained, by the way that present invention design is applied in the coding unit partition process of QTBT algorithms, can reduced by above-mentioned
Recursive execution is divided, reaches on the basis of video information complete transmission is ensured, greatly reduces the time of Video coding, is improved
Video coding treatment effeciency.
In addition, for the QTBT algorithms mentioned in present invention design, its QTBT default parameters in JEM4.0 is as follows
Shown in table 2：
Table 2
Also, in QTBT algorithms, its recurrence division rule is specially：When coding unit CU wide Width meets formula
(1) when, and binary tree divides depth B TDepth when meeting formula (2), can just carry out vertical binary tree division；When coding is single
When first CU high Height meets formula (3), and binary tree divides depth B TDepth when meeting formula (2), can just enter water-filling
Flat binary tree divides；Wherein, formula (1)~(4) are as follows：
minBTSize<Width≤maxBTSize (1)
BTDepth<maxBTDepth (2)
minBTSize<Height≤maxBTSize (3)
Width>minQTSize (4)
And once employ binary tree division, then quad-tree partition will not be used again afterwards.Wherein, the quad-tree partition
Condition be to meet formula (4), and for it need to carry out the block of quad-tree partition, the wide and high of coding unit CU blocks must be all
It is equal, so only needing to judge wide Width.
Above is the preferable implementation to the present invention is illustrated, but the invention is not limited to the implementation
Example, those skilled in the art can also make a variety of equivalent variations or replace on the premise of without prejudice to spirit of the invention
Change, these equivalent deformations or replacement are all contained in the application claim limited range.
Claims (10)
1. a kind of quick decision methods of QTBT based on KB wave filters, it is characterised in that：The step of this method, includes：
The first image is obtained, wherein, described first image refers to filtering video frame images using Ker-Bohme wave filters
The filtering image obtained after ripple processing；
According to the block size of current coded unit, corresponding high frequency points threshold value is chosen；
The high frequency points maximum of image-region is found out, wherein, described image region refers to compiling with current in the first image
The corresponding image-region of code unit；
Judge whether the high frequency points maximum found out is less than the high frequency points threshold value selected, if so, then jumping out present encoding list
Member divides recursive program；Conversely, then proceeding to divide recurrence to current coded unit according to QTBT algorithms.
2. quick decision methods of a kind of QTBT based on KB wave filters according to claim 1, it is characterised in that：It is described to obtain
The step for first image, it is specifically included：
Obtain expanding image after video frame images to be expanded to 1 row and 1 row in a mirror-image fashion；
The nuclear matrix of Ker-Bohme wave filters is obtained into filtering image, the obtained filtering with expanding after image carries out convolution
The first image that image is obtained for needed for.
3. quick decision methods of a kind of QTBT based on KB wave filters according to claim 2, it is characterised in that：The Ker-
The nuclear matrix KB of Bohme wave filters is as follows：
<mrow>
<mi>K</mi>
<mi>B</mi>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mn>4</mn>
</mfrac>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mrow>
<mo>-</mo>
<mn>1</mn>
</mrow>
</mtd>
<mtd>
<mn>2</mn>
</mtd>
<mtd>
<mrow>
<mo>-</mo>
<mn>1</mn>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>2</mn>
</mtd>
<mtd>
<mrow>
<mo>-</mo>
<mn>4</mn>
</mrow>
</mtd>
<mtd>
<mn>2</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>-</mo>
<mn>1</mn>
</mrow>
</mtd>
<mtd>
<mn>2</mn>
</mtd>
<mtd>
<mrow>
<mo>-</mo>
<mn>1</mn>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
4. according to a kind of quick decision methods of any one of claim 1-3 QTBT based on KB wave filters, it is characterised in that：
The step of this method, also includes the step for high frequency points threshold value is obtained, and the step for high frequency points threshold value is obtained includes
Have：
Multiple filtered samples figures are obtained after being filtered out respectively to multiple Sample video two field pictures using Ker-Bohme wave filters
Picture；
According to default block size, cutting division is carried out to multiple filtered samples images, so as to obtain multiple rectangular image blocks；
The lookup of high frequency points maximum is carried out to each rectangular image block, so as to find out the height in each rectangular image block
Frequency maximum；
According to the numerical value of the multiple high frequency points maximums found out, so as to be ranked up to multiple rectangular image blocks；
Query search is carried out to the rectangular image block after sequence by the way of binary chop, meets default so as to find out
The rectangular image block of condition；
It regard the high frequency points maximum in the rectangular image block found out as the high frequency points threshold value corresponding with default block size.
5. quick decision methods of a kind of QTBT based on KB wave filters according to claim 4, it is characterised in that：It is described default
Block size include 8x8,8x16,16x8,8x32,32x8,16x16,16x32,32x16,32x32,16x64,64x16,
32x64,64x32 and 64x64.
6. a kind of quick decision systems of QTBT based on KB wave filters, it is characterised in that：The system includes：
Acquisition module, for obtaining the first image, wherein, described first image refer to using Ker-Bohme wave filters to regarding
Frequency two field picture is filtered the filtering image obtained after processing；
Module is chosen, for the block size according to current coded unit, corresponding high frequency points threshold value is chosen；
Searching modul, the high frequency points maximum for finding out image-region, wherein, described image region is referred in the first figure
The image-region corresponding with current coded unit as in；
Whether judge module, the high frequency points maximum for judging to find out is less than the high frequency points threshold value selected, if so, then jumping
Go out current coded unit and divide recursive program；Conversely, then proceeding to divide recurrence to current coded unit according to QTBT algorithms.
7. quick decision systems of a kind of QTBT based on KB wave filters according to claim 6, it is characterised in that：It is described to obtain
Module is specifically included：
Image expands submodule, for obtaining expanding image after video frame images to be expanded to 1 row and 1 row in a mirror-image fashion；
Submodule is filtered, for the nuclear matrix of Ker-Bohme wave filters to be obtained into filtering image with expanding after image carries out convolution,
The first image that the obtained filtering image is obtained for needed for.
8. quick decision systems of a kind of QTBT based on KB wave filters according to claim 7, it is characterised in that：The Ker-
The nuclear matrix KB of Bohme wave filters is as follows：
<mrow>
<mi>K</mi>
<mi>B</mi>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mn>4</mn>
</mfrac>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mrow>
<mo>-</mo>
<mn>1</mn>
</mrow>
</mtd>
<mtd>
<mn>2</mn>
</mtd>
<mtd>
<mrow>
<mo>-</mo>
<mn>1</mn>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>2</mn>
</mtd>
<mtd>
<mrow>
<mo>-</mo>
<mn>4</mn>
</mrow>
</mtd>
<mtd>
<mn>2</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>-</mo>
<mn>1</mn>
</mrow>
</mtd>
<mtd>
<mn>2</mn>
</mtd>
<mtd>
<mrow>
<mo>-</mo>
<mn>1</mn>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
9. according to a kind of quick decision systems of any one of claim 6-8 QTBT based on KB wave filters, it is characterised in that：
The system also includes high frequency points threshold value acquisition module, and the high frequency points threshold value acquisition module is specifically included：
First acquisition submodule, for being filtered out respectively to multiple Sample video two field pictures using Ker-Bohme wave filters
After obtain multiple filtered samples images；
Second acquisition submodule, for according to default block size, cutting division to be carried out to multiple filtered samples images, so that
To multiple rectangular image blocks；
3rd acquisition submodule, for carrying out the lookup of high frequency points maximum to each rectangular image block, so as to find out every
High frequency points maximum in one rectangular image block；
Sorting sub-module, for the numerical value according to the multiple high frequency points maximums found out, so as to enter to multiple rectangular image blocks
Row sequence；
Submodule is searched, for carrying out query search to the rectangular image block after sequence by the way of binary chop, from
And find out the rectangular image block for meeting preparatory condition；
Threshold value determination sub-module, for the high frequency points maximum in the rectangular image block that will find out as with default block size
Corresponding high frequency points threshold value.
10. quick decision systems of a kind of QTBT based on KB wave filters according to claim 9, it is characterised in that：It is described pre-
If block size include 8x8,8x16,16x8,8x32,32x8,16x16,16x32,32x16,32x32,16x64,64x16,
32x64,64x32 and 64x64.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710229733.6A CN107197253B (en) | 2017-04-10 | 2017-04-10 | QTBT quick judgment method and system based on KB filter |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710229733.6A CN107197253B (en) | 2017-04-10 | 2017-04-10 | QTBT quick judgment method and system based on KB filter |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107197253A true CN107197253A (en) | 2017-09-22 |
| CN107197253B CN107197253B (en) | 2019-12-27 |
Family
ID=59871283
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710229733.6A Expired - Fee Related CN107197253B (en) | 2017-04-10 | 2017-04-10 | QTBT quick judgment method and system based on KB filter |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107197253B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108322745A (en) * | 2018-02-28 | 2018-07-24 | 中南大学 | Fast selecting method in a kind of frame based on inseparable quadratic transformation pattern |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040165765A1 (en) * | 2003-02-05 | 2004-08-26 | Jun-Ho Sung | Method and an appraratus to divide image blocks |
| CN104396261A (en) * | 2012-04-16 | 2015-03-04 | 三星电子株式会社 | Video coding method and device using high-speed edge detection, and related video decoding method and device |
| CN105430416A (en) * | 2015-12-04 | 2016-03-23 | 四川大学 | Fingerprint image compression method based on adaptive sparse domain coding |
| CN105721866A (en) * | 2016-02-01 | 2016-06-29 | 北京奇艺世纪科技有限公司 | Method and device for dividing coding unit |
-
2017
- 2017-04-10 CN CN201710229733.6A patent/CN107197253B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040165765A1 (en) * | 2003-02-05 | 2004-08-26 | Jun-Ho Sung | Method and an appraratus to divide image blocks |
| CN104396261A (en) * | 2012-04-16 | 2015-03-04 | 三星电子株式会社 | Video coding method and device using high-speed edge detection, and related video decoding method and device |
| CN105430416A (en) * | 2015-12-04 | 2016-03-23 | 四川大学 | Fingerprint image compression method based on adaptive sparse domain coding |
| CN105721866A (en) * | 2016-02-01 | 2016-06-29 | 北京奇艺世纪科技有限公司 | Method and device for dividing coding unit |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108322745A (en) * | 2018-02-28 | 2018-07-24 | 中南大学 | Fast selecting method in a kind of frame based on inseparable quadratic transformation pattern |
| CN108322745B (en) * | 2018-02-28 | 2019-12-03 | 中南大学 | Fast selecting method in a kind of frame based on inseparable quadratic transformation mode |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107197253B (en) | 2019-12-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| DE112017006657B4 (en) | Adaptive planar prediction with unequal weights | |
| CN103155557B (en) | Adaptive loop filter method and apparatus based on subregion basis | |
| CN103460699B (en) | Loop circuit filtering method and its equipment | |
| KR101838341B1 (en) | Depth coding | |
| KR20200136390A (en) | Image processing apparatus and image processing method | |
| CN107431813A (en) | Use the method and apparatus of the conversion process vision signal based on figure | |
| CN104754362B (en) | Image compression method using fine-divided block matching | |
| CN108353165A (en) | The method and apparatus that image is encoded/decoded using geometric modification picture | |
| SE1850159A2 (en) | Procedure for selecting a candidate block for fusion as well as a device for applying this procedure | |
| CN105850121A (en) | Method of filter control for block-based adaptive loop filtering | |
| CN108271024A (en) | Image coding, coding/decoding method and device | |
| Zhang et al. | Fast CU decision-making algorithm based on DenseNet network for VVC | |
| CN103051892B (en) | Embedded loop filter method and embedded loop filter | |
| CN109587491A (en) | A kind of intra-frame prediction method, device and storage medium | |
| CN113068031B (en) | Loop filtering method based on deep learning | |
| CN108737819B (en) | Flexible coding unit partitioning method based on quadtree binary tree structure | |
| CN117041599A (en) | HEVC-VPCC-based intra-frame rapid coding method and system | |
| CN107040778A (en) | Loop circuit filtering method and loop filter | |
| Hari et al. | Ctu partition for intra-mode hevc using convolutional neural network | |
| CN107197253A (en) | A kind of quick decision methods of QTBT based on KB wave filters and system | |
| KR20170053720A (en) | Method of lookup table size reduction for depth modelling mode in depth coding | |
| CN105979262B (en) | A kind of SAO optimization method of the HEVC based on temporal correlation and intra prediction direction | |
| CN109618169B (en) | Intra-frame decision method, device and storage medium for HEVC | |
| CN107343199B (en) | Rapid adaptive compensation method for sampling points in HEVC (high efficiency video coding) | |
| CN109040756B (en) | HEVC image content complexity-based rapid motion estimation method |
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: 20191227 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |