CN104506867B - Sample point self-adapted offset parameter method of estimation and device - Google Patents
Sample point self-adapted offset parameter method of estimation and device Download PDFInfo
- Publication number
- CN104506867B CN104506867B CN201410720721.XA CN201410720721A CN104506867B CN 104506867 B CN104506867 B CN 104506867B CN 201410720721 A CN201410720721 A CN 201410720721A CN 104506867 B CN104506867 B CN 104506867B
- Authority
- CN
- China
- Prior art keywords
- skipped
- sao
- columns
- line number
- component
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention discloses a kind of method for parameter estimation and device of point self-adapted skew of sampling, this method includes：GPU uses default number of threads, it is first determined the SAO statistical informations of all encoding blocks in current encoded frame；According to the dependence of the encoding block, decision-making treatment is carried out, the SAO parameters of all encoding blocks of the coded frame are obtained.Pass through the present invention, using GPU, accordingly counted by the pixel value to encoding block, complete the decision-making of the SAO parameters of each encoding block, the acceleration of the processing module is realized by means of GPU parallel processing capability, helps to realize the acceleration of the coding of CPU+GPU hybrid coding frameworks.
Description
Technical field
The present invention relates to technology of video compressing encoding field, in particular to the point self-adapted skew of one kind sampling
The method for parameter estimation and device of (Sample Adaptive Offset, referred to as SAO).
Background technology
Efficient video coding (High Efficiency Video Coding, referred to as HEVC) is by International Standards Organization
(International Organization for Standardization, referred to as ISO) and International Telecommunication Association (ITU
Telecommunication Standardization Sector, referred to as ITU-T) combine the Video coding of new generation proposed
Standard, the standard can realize higher video image compression efficiency relative to conventional coding standard.HEVC supports Main,
Three kinds of class of Main10 and still image, support YUV4:2:The Signal coding of 0 sampling form, compresses effect compared with H.264/AVC
Rate is doubled.However, because HEVC algorithm complex is high, code rate is relatively low, it is difficult to be realized on PC real
When coding, thus have many researchers start using graphics processor (Graphic Processing Unit, referred to as
GPU the optimization of encoder) is carried out, the coding of high-speed is realized whereby and preferable effect is achieved.For example for HEVC's
Main movement estimation module GPU optimization, and deblocking filter GPU accelerating algorithms.
The point self-adapted skew (Sample Adaptive Offset, referred to as SAO) of sampling is emerging one in HEVC
Filtering algorithm is planted, its algorithm complex is not high, the time ratio taken in the coding time of reference software is less than 1%, temporarily at present
When without disclosed GPU accelerating algorithms.But due to the particularity of its processing procedure so that the algorithm is difficult by single instruction multiple
Optimized according to instruction.Therefore, in the optimization process to encoder, it is necessary to also carry out GPU optimizations to SAO.SAO passes through inspection
The boundary direction and pixel number of the reconstruction frames of coded image are surveyed, by different with the point application of pixel range to different directions
Offset, to cause reconstruction frames image and the image to be encoded is more nearly, thus bring preferable performance gain.
In HEVC coded treatment, the encoding block of constant size can be divided the image into, for example its size is configurable to
32x32.In SAO processing, to each encoding block, an optimal decision-making classification and its corresponding four compensation can be distributed
Value.SAO type be divided into boundary compensation (Edge Offset, referred to as EO) and band compensation (Band Offset, referred to as
BO) two major class.In EO, according to the direction on border, it is divided into 0 degree, 45 degree, 90 degree and 135 degree four classifications.Band in BO is
Band, is a kind of mapping of pixel value, such as the image of 8 locating depths, and pixel value span is 0~255, any pixel value x's
Band is numberedDue to the centrality of color, in band compensation, continuous 4 bands are generally only taken, because
And the band numbering that an extra numerical value carrys out record start is needed, in addition, for each band, encoder can be needed
The records of values of compensation gets off and is encoded in code stream.The SAO example of parameters of one encoding block is as shown in table 1.Wherein refer to
Pattern is set to without reference to left side encoding block and just has the coding of information below during without reference to top encoding block, and only SAO is beaten
When opening, SAO types and corresponding offset just need to be encoded.
The SAO parameter formats of 1 one encoding blocks of table
For the low above mentioned problem of the efficiency comparison of SAO method for parameter estimation in correlation technique, not yet propose at present effective
Solution.
The content of the invention
For SAO method for parameter estimation in correlation technique efficiency comparison it is low the problem of, the invention provides a kind of SAO ginsengs
Number estimation method and device, at least to solve the problem.
On the one hand there is provided the point self-adapted skew SAO method for parameter estimation of one kind sampling, including：Graphics processor GPU makes
Default number of threads is used, the SAO parameters of the code segment block in current encoded frame are determined；Closed according to the dependence of the encoding block
System, carries out decision-making treatment, obtains the SAO parameters of all encoding blocks of the coded frame.
Preferably, the GPU determines the SAO parameters of the code segment block in current encoded frame using 1 sets of threads warp
Including：Each thread in 1 sets of threads warp in the GPU corresponds to each thread the default position of the coded frame
The numerical value for the element put copies caching to, wherein, the pixel numerical value of each coding block copy is described one in the coded frame
The integral multiple of individual sets of threads warp sizes；In the caching, each thread determines the corresponding row picture of each thread
The difference of the boundary types of element value and the pixel value of the corresponding original image frame；The row excluded according to preset rules, by institute
State and arrange the information of corresponding thread and be set to zero；According to one sets of threads warp statistical information, Boundary Statistic EO is determined
Information and band statistics BO information.
Preferably, the EO information includes：16 groups of N and E values of one color component, the BO information includes：32 groups of N and
E values.
Preferably, the points that different colours component is skipped include：
When SAO parameter types are 0 degree of direction of EO, it is 4/3 that line number is skipped in bottom in luminance component, and row are skipped on the right
Number is 5；It is 2 that line number is skipped in bottom in chromatic component, and it is 3 that columns is skipped on the right；
When SAO parameter types are 90 degree of directions of EO, it is 4 that line number is skipped in bottom in luminance component, and columns is skipped on the right
For 5/4；It is 2 that line number is skipped in bottom in chromatic component, and it is 3/2 that columns is skipped on the right；
When SAO parameter types are 135 degree of directions of EO, it is 4 that line number is skipped in bottom in luminance component, and row are skipped on the right
Number is 5；It is 2 that line number is skipped in bottom in chromatic component, and it is 3 that columns is skipped on the right；
When SAO parameter types are 45 degree of directions of EO, it is 4 that line number is skipped in bottom in luminance component, and columns is skipped on the right
For 5；It is 2 that line number is skipped in bottom in chromatic component, and it is 3 that columns is skipped on the right；
When SAO parameter types are BO, it is 4/3 that line number is skipped in bottom in luminance component, and it is 5/4 that columns is skipped on the right；
It is 2/1 that line number is skipped in bottom in chromatic component, and it is 3/2 that columns is skipped on the right；
Numeral wherein behind slash is to use the points skipped before block elimination filtering during pixel numerical value.
Preferably, according to the dependence of the encoding block, decision-making treatment is carried out, all codings of the coded frame are obtained
The SAO parameters of block include：
According to avoiding the peak hour, parallel form carries out the decision-making treatment.
On the other hand, a kind of point self-adapted skew SAO parameter estimation apparatus of sampling is additionally provided, applied to GPU, including：
Determining module, for using default number of threads, determining the SAO parameters of the code segment block in current encoded frame；Processing module,
For the dependence according to the encoding block, decision-making treatment is carried out, the SAO ginsengs of all encoding blocks of the coded frame are obtained
Number.
Preferably, the determining module includes：
First determining unit, when using 1 sets of threads warp for the GPU, for each thread to be corresponded into institute
The numerical value for stating the element of the predeterminated position of coded frame copies caching to, wherein, the picture of each coding block copy in the coded frame
Vegetarian refreshments numerical value is the integral multiple of one sets of threads warp sizes；
Second determining unit, in the caching, each thread determines the corresponding row picture of each thread
The difference of the boundary types of element value and the pixel value of the corresponding original image frame；
Setting unit, for the row excluded according to preset rules, zero is set to by the information of the corresponding thread of the row；
3rd determining unit, for the statistical information according to one sets of threads warp, determines Boundary Statistic EO information
BO information is counted with band.
Preferably, the EO information includes：16 groups of N and E values of one color component, the BO information includes：32 groups of N and
E values.
Preferably, the points that the setup module is skipped in different colours component include：
When SAO parameter types are 0 degree of direction of EO, it is 4/3 that line number is skipped in bottom in luminance component, and row are skipped on the right
Number is 5；It is 2 that line number is skipped in bottom in chromatic component, and it is 3 that columns is skipped on the right；
When SAO parameter types are 90 degree of directions of EO, it is 4 that line number is skipped in bottom in luminance component, and columns is skipped on the right
For 5/4；It is 2 that line number is skipped in bottom in chromatic component, and it is 3/2 that columns is skipped on the right；
When SAO parameter types are 135 degree of directions of EO, it is 4 that line number is skipped in bottom in luminance component, and row are skipped on the right
Number is 5；It is 2 that line number is skipped in bottom in chromatic component, and it is 3 that columns is skipped on the right；
When SAO parameter types are 45 degree of directions of EO, it is 4 that line number is skipped in bottom in luminance component, and columns is skipped on the right
For 5；It is 2 that line number is skipped in bottom in chromatic component, and it is 3 that columns is skipped on the right；
When SAO parameter types are BO, it is 4/3 that line number is skipped in bottom in luminance component, and it is 5/4 that columns is skipped on the right；
It is 2/1 that line number is skipped in bottom in chromatic component, and it is 3/2 that columns is skipped on the right；
Numeral wherein behind slash is to use the points skipped before block elimination filtering during pixel numerical value.
Preferably, the processing module carries out the decision-making treatment according to parallel form of avoiding the peak hour.
By the present invention, using GPU, accordingly counted by the pixel value to encoding block, complete each encoding block
The decision-making of SAO parameters, the acceleration of the processing module is realized by means of GPU parallel processing capability, and help realizes that CPU+GPU is mixed
The acceleration of the coding of coding scheme.
Brief description of the drawings
Accompanying drawing described herein is used for providing a further understanding of the present invention, constitutes the part of the application, this hair
Bright schematic description and description is used to explain the present invention, does not constitute inappropriate limitation of the present invention.In the accompanying drawings：
Fig. 1 is the flow chart of SAO method for parameter estimation according to embodiments of the present invention；
Fig. 2 is the schematic diagram of the SAO decision sequences of different masses according to embodiments of the present invention；
Fig. 3 is the flow chart of the SAO Parameter Decision Making flows of an encoding block according to embodiments of the present invention；
Fig. 4 is the structured flowchart of SAO parameter estimation apparatus according to embodiments of the present invention.
Embodiment
Describe the present invention in detail below with reference to accompanying drawing and in conjunction with the embodiments.It should be noted that not conflicting
In the case of, the feature in embodiment and embodiment in the application can be mutually combined.
Fig. 1 is the flow chart of SAO method for parameter estimation according to embodiments of the present invention, and such as Fig. 1 is somebody's turn to do, and this method includes as follows
The step of：
Step S102, GPU determine the SAO statistics letters of all encoding blocks in current encoded frame using default number of threads
Breath.
Step S104, according to the dependence of the encoding block, carries out decision-making treatment, obtains all codings of current encoded frame
The SAO parameters of block.
In correlation technique, the estimation of SAO parameters uses serial ground mode traversal mode, and its is relatively inefficient, leads to
Above-mentioned steps are crossed, are handled with carrying out parallelization using GPU, can be same on the premise of parameter relationship dependency between keeping encoding block
When estimate the SAO parameters of multiple encoding blocks, and the also multiple compensation of parallel computation when carrying out parameter Estimation to each encoding block
Value parameter configuration is corresponding to be spent, and improves the degree of parallelism of algorithm so that the time complexity of SAO parameter Estimations is substantially reduced.
Preferably, the GPU can use 1 sets of threads warp, above-mentioned steps S102 to perform in the following way：
Each thread is corresponded to the pre- of the coded frame by each thread in 1 sets of threads warp in the GPU
If the numerical value of the element of position copies caching to, wherein, the pixel numerical value of each coding block copy is institute in the coded frame
State the integral multiple of a sets of threads warp size；
In the caching, each thread determines the boundary types of the corresponding row pixel value of each thread and right
The difference of the pixel value for the original image frame answered；
The row excluded according to preset rules, zero is set to by the information of the corresponding thread of the row；
According to one sets of threads warp statistical information, the SAO statistical informations of all colours component are determined.
Preferably, the SAO information of one color component includes the EO information and BO information of the color component, wherein EO
Information includes：16 groups of N and E values of the color component, the BO information includes：32 groups of N and E values of the color component, wherein N is
The pixel quantity of particular category, picture when E performs for the primitive frame pixel value of the pixel position of particular category with the module
The absolute value sum of the difference of element value.
As another preferably embodiment, the points that different colours component is skipped include：
When SAO parameter types are 0 degree of direction of EO, it is 4/3 that line number is skipped in bottom in luminance component, and row are skipped on the right
Number is 5；It is 2 that line number is skipped in bottom in chromatic component, and it is 3 that columns is skipped on the right；
When SAO parameter types are 90 degree of directions of EO, it is 4 that line number is skipped in bottom in luminance component, and columns is skipped on the right
For 5/4；It is 2 that line number is skipped in bottom in chromatic component, and it is 3/2 that columns is skipped on the right；
When SAO parameter types are 135 degree of directions of EO, it is 4 that line number is skipped in bottom in luminance component, and row are skipped on the right
Number is 5；It is 2 that line number is skipped in bottom in chromatic component, and it is 3 that columns is skipped on the right；
When SAO parameter types are 45 degree of directions of EO, it is 4 that line number is skipped in bottom in luminance component, and columns is skipped on the right
For 5；It is 2 that line number is skipped in bottom in chromatic component, and it is 3 that columns is skipped on the right；
When SAO parameter types are BO, it is 4/3 that line number is skipped in bottom in luminance component, and it is 5/4 that columns is skipped on the right；
It is 2/1 that line number is skipped in bottom in chromatic component, and it is 3/2 that columns is skipped on the right；
Numeral wherein behind slash is to use the points skipped before block elimination filtering during pixel numerical value.
As another preferably embodiment, step S104 can be carried out at the decision-making according to parallel form of avoiding the peak hour
Reason.
Illustrated below in conjunction with preferred embodiment, preferred embodiment below combines above-described embodiment and is preferable to carry out
Mode.
Preferred embodiment one
This preferred embodiment provides the SAO method for parameter estimation on a kind of GPU.In the preferred embodiment, from given
Reconstruction frames and original image frame pixel value before encoder filtering carry out the estimation of SAO parameters, are handled using GPU efficient parallel
Characteristic, first to each encoding block parallel computation statistical information, then carries out parallel decision-making, so that real to block according to dependence
The now acceleration at double of the module.It is further explained below：
In the preferred embodiment, the thread allocation strategy on GPU can include：
A) calculated in SAO statistical informations with SAO Parameter Decision Makings, each encoding block allocation number of threads is the one of GPU
Individual sets of threads warp size.
B) to each encoding block, often the pixel quantity of row copy is the integral multiple of a warp size, not additional copy
More than the pixel of block boundary.
Preferably, in the preferred embodiment, the partial information of image boundary is given up, to realize quick Information Statistics,
The pixel quantity that should be skipped to each encoding block is as shown in Table 2.
The points that the different colours component of table 2 is skipped
It should be noted that the numeral after slash is to use the point that should be skipped during pixel numerical value before block elimination filtering
Number.
Preferred embodiment two
This preferred embodiment provides the quick SAO decision-making techniques on a kind of GPU.In the preferred embodiment, this method
SAO statistical informations can be included to calculate and two steps of SAO Parameter Decision Makings.Give full play to GPU's when SAO statistical informations are calculated
Parallel processing capability, can simultaneously be handled with whole two field picture；In SAO Parameter Decision Makings, it is limited to the dependence between block, according to mistake
The mode of peak parallel processing is handled.
As a preferably embodiment, it is not present to rely in the calculating of SAO statistical informations, between each encoding block and closes
System, individually can be handled by a sets of threads, if sets of threads quantity is enough can to handle all blocks simultaneously.
Preferably, if in advance carry out SAO statistical information calculating, the number of blocks that can be handled simultaneously be current reconstructed frame
The block of generation；
Preferably, when carrying out SAO Parameter Decision Makings, parallel processing is carried out using parallel processing technique of avoiding the peak hour, is fully improved
Degree of parallelism.
Preferred embodiment two
This preferred embodiment determines method there is provided a kind of snapshot relation of net list compartment.In the preferred embodiment,
Using GPU powerful parallel processing capability, piecemeal is based on to coded image and carries out the SAO Information Statistics of fast parallel ground, Ran Hou
Ensure to perform quick SAO Parameter Decision Makings under conditions of dependence, ensureing condition of the objective evaluation quality without significantly sacrificing
The acceleration of lower realization at double.
Illustrated below by embodiment.
This preferred embodiment includes：SAO statistical informations are calculated.
Specifically, in the processing that SAO is filtered, non-borderline pixel can all be classified and then filter in encoding block,
The offset of each classification, the difference (being designated as E) dependent on the value and original image of the quantity (being designated as N) of such point and such point.
In EO four classifications, each classification divide into four subclass again, thus have 16 groups of N, E values to a color component.And BO
Establishing criteria has 32 groups, so have 32 groups of N, E values.Therefore each color component is needed to calculate 48 pairs of N, E values.Boundary Statistic is believed
In the calculating of breath and band statistical information, it is that each encoding block (size is NxN) distributes a warp, thus avoids extra same
Step.
Preferably, the statistics to boundary information is carried out in accordance with the following steps：
The first step：The each thread of sets of threads copies correspondence position element numerical value to shared buffer memory.
Second step：The last column limited since the row of SAO filtering process into form 2, each thread calculates correspondence
The boundary types of one row pixel value and corresponding and original image difference.
3rd step：The row that should be excluded according to 2 pairs of table, the local statistic information of the corresponding thread of respective column is reset.
4th step：The statistical information of cumulative sets of threads, obtains N, E value i.e. statistical information of 4 subclass.
Preferably, in BO statistical information is calculated, sets of threads is divided into 4 groups according to thread number, and each group is complete
Into the Information Statistics of one part of pixel value.Wherein, different colours component and SAO types need the boundary point for the encoding block skipped
Quantity is as shown in Table 2.
This preferred embodiment includes：SAO Parameter Decision Makings.
Preferably, similar to other code elements, the SAO parameters of an encoding block can also refer to its top or a left side
The parameter of the block on side, this point brings dependence between encoding block, on the premise of this dependence is not destroyed, can be with
The SAO parameter Estimations of all encoding blocks are completed with such as graph traversal order, wherein the block for indicating identical numerical value is can be same
Time-triggered protocol, and numerical value the greater needs to wait less piece of numerical value to be disposed.
Under this framework, it is that block in each decision-making distributes a sets of threads, carries out decision-making treatment, its handling process
It is：
The first step：Initialization, resets the context model of all rows, and the parameter array with reference to block is set to sky, next time
The position transverse and longitudinal coordinate sum of process block is set to seq=0.
Second step：To the block that all transverse and longitudinal coordinate sums are seq, note c is the state of context model, according to as shown in the figure
Flow, perform SAO parameter Estimations；
3rd step：By all pieces of the SAO parameter read-ins overall situation storage of current decision, and use shared buffer memory backup event
The follow-up reference of meaning；
4th step：If all pieces are disposed, terminate, otherwise continued cycling through after seq increases by 1 from second step.
It should be noted that can be in such as one group computer executable instructions the step of the flow of accompanying drawing is illustrated
Performed in computer system, and, although logical order is shown in flow charts, but in some cases, can be with not
The order being same as herein performs shown or described step.
As used below, term " submodule ", " module " can realize the software of predetermined function and/or the group of hardware
Close.Although the system and method described by following examples are preferably realized with software, hardware, or software and hardware
The realization of combination be also that may and be contemplated.
In another embodiment, a kind of SAO Parameter Estimation Softwares are additionally provided, the software is used to perform above-mentioned implementation
Technical scheme described in example and preferred embodiment.
In another embodiment, a kind of storage medium is additionally provided, be stored with above-mentioned SAO parameters in the storage medium
Estimation Software, the storage medium includes but is not limited to：CD, floppy disk, hard disk, scratch pad memory etc..
The embodiment of the present invention additionally provides a kind of SAO parameter estimation apparatus, and the SAO parameter estimation apparatus can be used for realizing
Above-mentioned SAO method for parameter estimation and preferred embodiment, had carried out explanation, had repeated no more, below to SAO parameter Estimations
The module being related in device is illustrated.As used below, term " module " can realize predetermined function software and/
Or the combination of hardware.Although the system and method described by following examples are realized preferably with software, hardware, or soft
The realization of the combination of part and hardware is also that may and be contemplated.
Fig. 4 is the structured flowchart of SAO parameter estimation apparatus according to embodiments of the present invention, and the device can apply to GPU,
As shown in figure 4, the device includes：Determining module 42, processing module 44, said structure is described in detail below.
Determining module 42, for using default number of threads, determining the SAO statistics of all encoding blocks in current encoded frame
Information；Processing module 44, is connected to determining module 42, for the dependence according to the encoding block, carries out decision-making treatment, obtains
To the SAO parameters of all encoding blocks of the coded frame.
Preferably, the determining module can include：
First determining unit, when using 1 sets of threads warp for the GPU, for each thread to be corresponded into institute
The numerical value for stating the element of the predeterminated position of coded frame copies caching to, wherein, the picture of each coding block copy in the coded frame
Vegetarian refreshments numerical value is the integral multiple of one sets of threads warp sizes；
Second determining unit, in the caching, each thread determines the corresponding row picture of each thread
The difference of the boundary types of element value and the pixel value of the corresponding original image frame；
Setting unit, for the row excluded according to preset rules, zero is set to by the information of the corresponding thread of the row；
3rd determining unit, for the statistical information according to one sets of threads warp, determines Boundary Statistic EO information
BO information is counted with band.
Preferably, the EO information includes：16 groups of N and E values of one color component, the BO information includes：32 groups of N and
E values.
Preferably, the points that the setup module is skipped in different colours component include：
When SAO parameter types are 0 degree of direction of EO, it is 4/3 that line number is skipped in bottom in luminance component, and row are skipped on the right
Number is 5；It is 2 that line number is skipped in bottom in chromatic component, and it is 3 that columns is skipped on the right；
When SAO parameter types are 90 degree of directions of EO, it is 4 that line number is skipped in bottom in luminance component, and columns is skipped on the right
For 5/4；It is 2 that line number is skipped in bottom in chromatic component, and it is 3/2 that columns is skipped on the right；
When SAO parameter types are 135 degree of directions of EO, it is 4 that line number is skipped in bottom in luminance component, and row are skipped on the right
Number is 5；It is 2 that line number is skipped in bottom in chromatic component, and it is 3 that columns is skipped on the right；
When SAO parameter types are 45 degree of directions of EO, it is 4 that line number is skipped in bottom in luminance component, and columns is skipped on the right
For 5；It is 2 that line number is skipped in bottom in chromatic component, and it is 3 that columns is skipped on the right；
When SAO parameter types are BO, it is 4/3 that line number is skipped in bottom in luminance component, and it is 5/4 that columns is skipped on the right；
It is 2/1 that line number is skipped in bottom in chromatic component, and it is 3/2 that columns is skipped on the right；
Numeral wherein behind slash is to use the points skipped before block elimination filtering during pixel numerical value.
Preferably, the processing module carries out the decision-making treatment according to parallel form of avoiding the peak hour.
By above-described embodiment there is provided a kind of SAO method for parameter estimation and device, the SAO parameter Estimation sides on GPU
Method.Its processing procedure is divided into SAO information and calculated and two parts of SAO Parameter Decision Makings.According to GPU treatment characteristic, suitably pacify
The workload of thread and algorithm process flow on GPU are arranged, higher speed-up ratio can be realized.It is fully sharp in the calculating of SAO information
With GPU computing capability, while multiple pieces are handled, and in the processing of a block, it is to avoid extra single pixel value
Copy, thus lose very little in the case of realize higher speed-up ratio, at the same time, it is proposed that should before block elimination filtering
The boundary point information skipped.In SAO decision-makings, the dependence of the context coding model between row is eliminated, it is parallel using avoiding the peak hour
Higher degree of parallelism is realized in processing.
Have it should be noted that these technique effects are not above-mentioned all embodiments, some technologies effect
The some preferred embodiments of fruit could be obtained.
Obviously, those skilled in the art should be understood that above-mentioned each module of the invention or each step can be with general
Computing device realize that they can be concentrated on single computing device, or be distributed in multiple computing devices and constituted
Network on, alternatively, the program code that they can be can perform with computing device be realized, so as to being stored in
Performed in storage device by computing device, they are either fabricated to each integrated circuit modules respectively or by them
Multiple modules or step single integrated circuit module is fabricated to realize.So, the present invention is not restricted to any specific hard
Part and software are combined.
The preferred embodiments of the present invention are the foregoing is only, are not intended to limit the invention, for the skill of this area
For art personnel, the present invention can have various modifications and variations.Within the spirit and principles of the invention, that is made any repaiies
Change, equivalent substitution, improvement etc., should be included in the scope of the protection.
Claims (8)
1. the point self-adapted skew SAO method for parameter estimation of one kind sampling, it is characterised in that including：
Graphics processor GPU, according to the dependence of encoding block, determines the coding in current encoded frame using default number of threads
The SAO parameters of block；
GPU determines the SAO parameters of one or more of current encoded frame encoding block using a sets of threads warp, including：
Each thread in a sets of threads warp in GPU corresponds to each thread the element of the predeterminated position of coded frame
Numerical value copy caching to, wherein, the pixel numerical value of each coding block copy is a sets of threads warp size in coded frame
Integral multiple；
In the buffer, each thread determines the boundary types and corresponding original image frame of the thread corresponding row pixel value
The difference of pixel value；
The row excluded according to preset rules, zero is set to by the information for arranging corresponding thread；
According to sets of threads warp statistical information, the Boundary Statistic EO information and band statistics BO of each color component are determined
Information.
2. a kind of point self-adapted skew SAO method for parameter estimation of sampling according to claim 1, it is characterised in that EO believes
Breath includes：16 groups of N and E values, BO information includes：32 groups of N and E values.
3. a kind of point self-adapted skew SAO method for parameter estimation of sampling according to claim 2, it is characterised in that different
The points that color component is skipped include：
When SAO parameter types are 0 degree of direction of EO, it is 4/3 that line number is skipped in bottom in luminance component, and the right skips columns and is
5；It is 2 that line number is skipped in bottom in chromatic component, and it is 3 that columns is skipped on the right；
When SAO parameter types are 90 degree of directions of EO, it is 4 that line number is skipped in bottom in luminance component, and it is 5/ that columns is skipped on the right
4；It is 2 that line number is skipped in bottom in chromatic component, and it is 3/2 that columns is skipped on the right；
When SAO parameter types are 135 degree of directions of EO, it is 4 that line number is skipped in bottom in luminance component, and the right skips columns and is
5；It is 2 that line number is skipped in bottom in chromatic component, and it is 3 that columns is skipped on the right；
When SAO parameter types are 45 degree of directions of EO, it is 4 that line number is skipped in bottom in luminance component, and it is 5 that columns is skipped on the right；
It is 2 that line number is skipped in bottom in chromatic component, and it is 3 that columns is skipped on the right；
When SAO parameter types are BO, it is 4/3 that line number is skipped in bottom in luminance component, and it is 5/4 that columns is skipped on the right；In color
It is 2/1 that line number is skipped in bottom in degree component, and it is 3/2 that columns is skipped on the right；
Numeral wherein behind slash is to use the points skipped before block elimination filtering during pixel numerical value.
4. a kind of point self-adapted skew SAO method for parameter estimation of sampling according to any one of claim 1 to 3, it is special
Levy and be, according to the dependence of encoding block, carry out decision-making treatment, obtaining the SAO parameters of all encoding blocks of coded frame includes：
According to avoiding the peak hour, parallel form carries out decision-making treatment.
5. the point self-adapted skew SAO parameter estimation apparatus of one kind sampling, applied to image processor GPU, it is characterised in that bag
Include：
Determining module, for using default number of threads, determining the SAO information of the encoding block in current encoded frame；
Processing module, for the dependence according to encoding block, carries out decision-making treatment, obtains the SAO ginsengs of the encoding block of coded frame
Number；
Determining module includes：First determining unit, when using a sets of threads warp for GPU, for by each thread pair
Caching should be copied in the numerical value of the element of the predeterminated position of coded frame, wherein, the pixel of each coding block copy in coded frame
Point value is the integral multiple of a sets of threads warp size；
Second determining unit, in the buffer, each thread determines the border class of the corresponding row pixel value of each thread
The difference of the pixel value of type and corresponding original image frame；
Setting unit, for the row excluded according to preset rules, zero is set to by the information for arranging corresponding thread；
3rd determining unit, for the statistical information according to a sets of threads warp, determines the Boundary Statistic of each color component
EO information and band statistics BO information.
6. a kind of point self-adapted skew SAO parameter estimation apparatus of sampling according to claim 5, it is characterised in that EO believes
Breath includes：16 groups of N and E values, BO information includes：32 groups of N and E values.
7. a kind of point self-adapted skew SAO parameter estimation apparatus of sampling according to claim 6, it is characterised in that it is determined that
The points that module is skipped in different colours component include：
When SAO parameter types are 0 degree of direction of EO, it is 4/3 that line number is skipped in bottom in luminance component, and the right skips columns and is
5；It is 2 that line number is skipped in bottom in chromatic component, and it is 3 that columns is skipped on the right；
When SAO parameter types are 90 degree of directions of EO, it is 4 that line number is skipped in bottom in luminance component, and it is 5/ that columns is skipped on the right
4；It is 2 that line number is skipped in bottom in chromatic component, and it is 3/2 that columns is skipped on the right；
When SAO parameter types are 135 degree of directions of EO, it is 4 that line number is skipped in bottom in luminance component, and the right skips columns and is
5；It is 2 that line number is skipped in bottom in chromatic component, and it is 3 that columns is skipped on the right；
When SAO parameter types are 45 degree of directions of EO, it is 4 that line number is skipped in bottom in luminance component, and it is 5 that columns is skipped on the right；
It is 2 that line number is skipped in bottom in chromatic component, and it is 3 that columns is skipped on the right；
When SAO parameter types are BO, it is 4/3 that line number is skipped in bottom in luminance component, and it is 5/4 that columns is skipped on the right；In color
It is 2/1 that line number is skipped in bottom in degree component, and it is 3/2 that columns is skipped on the right；
Numeral wherein behind slash is to use the points skipped before block elimination filtering during pixel numerical value.
The point self-adapted skew SAO parameter estimation apparatus 8. one kind according to any one of claim 5 to 7 is sampled, it is special
Levy and be, processing module carries out decision-making treatment according to parallel form of avoiding the peak hour.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410720721.XA CN104506867B (en) | 2014-12-01 | 2014-12-01 | Sample point self-adapted offset parameter method of estimation and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410720721.XA CN104506867B (en) | 2014-12-01 | 2014-12-01 | Sample point self-adapted offset parameter method of estimation and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104506867A CN104506867A (en) | 2015-04-08 |
| CN104506867B true CN104506867B (en) | 2017-07-21 |
Family
ID=52948574
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410720721.XA Active CN104506867B (en) | 2014-12-01 | 2014-12-01 | Sample point self-adapted offset parameter method of estimation and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104506867B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105049853A (en) * | 2015-07-06 | 2015-11-11 | 深圳市云宙多媒体技术有限公司 | SAO coding method and system based on fragment source analysis |
| JP6634936B2 (en) * | 2016-04-05 | 2020-01-22 | 富士通株式会社 | Image encoding device, image encoding method, image encoding program, image decoding device, image decoding method, image decoding program, and image transmission system |
| CN105959695B (en) * | 2016-04-28 | 2019-07-09 | 中山大学 | A kind of adaptive offset method of parallel pixel based on GPU |
| CN114995898B (en) * | 2022-05-13 | 2024-08-13 | Oppo广东移动通信有限公司 | Target object processing method and device and electronic equipment |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102075753A (en) * | 2011-01-13 | 2011-05-25 | 中国科学院计算技术研究所 | Method for deblocking filtration in video coding and decoding |
| CN102835104A (en) * | 2010-04-14 | 2012-12-19 | 西门子企业通讯有限责任两合公司 | Method for deblocking filtering |
| CN103702126A (en) * | 2013-12-10 | 2014-04-02 | 清华大学深圳研究生院 | Parallel encoding optimization method based on standard video HEVC (High Efficiency Video Coding) |
| CN103956991A (en) * | 2014-04-10 | 2014-07-30 | 北京遥测技术研究所 | FIR filter parallel realization method based on CPU/GPU heterogeneous platform |
| CN103974081A (en) * | 2014-05-08 | 2014-08-06 | 杭州同尊信息技术有限公司 | HEVC coding method based on multi-core processor Tilera |
| CN104012091A (en) * | 2011-11-04 | 2014-08-27 | 高通股份有限公司 | Adaptive center band offset filter for video coding |
| CN104125466A (en) * | 2014-07-10 | 2014-10-29 | 中山大学 | GPU (Graphics Processing Unit)-based HEVC (High Efficiency Video Coding) parallel decoding method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9426469B2 (en) * | 2012-12-17 | 2016-08-23 | Broadcom Corporation | Combination HEVC deblocker/SAO filter |
-
2014
- 2014-12-01 CN CN201410720721.XA patent/CN104506867B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102835104A (en) * | 2010-04-14 | 2012-12-19 | 西门子企业通讯有限责任两合公司 | Method for deblocking filtering |
| CN102075753A (en) * | 2011-01-13 | 2011-05-25 | 中国科学院计算技术研究所 | Method for deblocking filtration in video coding and decoding |
| CN104012091A (en) * | 2011-11-04 | 2014-08-27 | 高通股份有限公司 | Adaptive center band offset filter for video coding |
| CN103702126A (en) * | 2013-12-10 | 2014-04-02 | 清华大学深圳研究生院 | Parallel encoding optimization method based on standard video HEVC (High Efficiency Video Coding) |
| CN103956991A (en) * | 2014-04-10 | 2014-07-30 | 北京遥测技术研究所 | FIR filter parallel realization method based on CPU/GPU heterogeneous platform |
| CN103974081A (en) * | 2014-05-08 | 2014-08-06 | 杭州同尊信息技术有限公司 | HEVC coding method based on multi-core processor Tilera |
| CN104125466A (en) * | 2014-07-10 | 2014-10-29 | 中山大学 | GPU (Graphics Processing Unit)-based HEVC (High Efficiency Video Coding) parallel decoding method |
Non-Patent Citations (1)
| Title |
|---|
| 基于HEVC视频编码标准的后处理技术的研究;盛希;《中国优秀硕士学位论文全文数据库 信息科技辑 》;20140315(第03期);正文第35-45页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104506867A (en) | 2015-04-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR100242636B1 (en) | Signal adaptive post processing system for reducing blocking effect and ringing noise | |
| CN104506867B (en) | Sample point self-adapted offset parameter method of estimation and device | |
| US20030020835A1 (en) | Methods and apparatus for removing compression artifacts in video sequences | |
| DE112009004320T5 (en) | Memory Subsystem | |
| CN1260949C (en) | Method and apparatus for improving picture sharpness | |
| US9179162B2 (en) | Image transform zero coefficient selection and zero-skip transmission for arbitrary shape transform coding | |
| CN102630011A (en) | Compressive perceptual coding and decoding method and system in video sensor network | |
| US7092580B2 (en) | System and method using edge processing to remove blocking artifacts from decompressed images | |
| CN110430403A (en) | A kind of image processing method and device | |
| CN110415188B (en) | HDR image tone mapping method based on multi-scale morphology | |
| KR100675498B1 (en) | Device and method for filtering | |
| US20040085315A1 (en) | Texture partition and transmission method for network progressive transmission and real-time rendering by using the wavelet coding algorithm | |
| CN106296593B (en) | Image restoration method and device | |
| EP0443114A2 (en) | Method and apparatus for image-data decompression | |
| US11600026B2 (en) | Data processing systems | |
| CN117560511A (en) | Spacer image compression method and system based on graph segmentation technology and electric power inspection | |
| Lakhani | Improved equations for JPEG's blocking artifacts reduction approach | |
| CN117196940A (en) | Super-resolution reconstruction method suitable for real scene image based on convolutional neural network | |
| CN1156166C (en) | Method and apparatus for determining optimum grid for use block-based video signal coding system | |
| CN109963158A (en) | A kind of high definition video decoding method based on GPU parallel computation | |
| CN114666590A (en) | Full-view video coding method and system based on load balancing | |
| CN106157252B (en) | A kind of image processing method and image processing apparatus | |
| CN112584154A (en) | Video coding block dividing method and device, electronic equipment and storage medium | |
| TW201419864A (en) | System and method for splitting an image | |
| US20090214131A1 (en) | System and method for multi-scale sigma filtering using quadrature mirror filters |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |