CN104243991B - A kind of side information generation method and device - Google Patents
A kind of side information generation method and device Download PDFInfo
- Publication number
- CN104243991B CN104243991B CN201410536223.XA CN201410536223A CN104243991B CN 104243991 B CN104243991 B CN 104243991B CN 201410536223 A CN201410536223 A CN 201410536223A CN 104243991 B CN104243991 B CN 104243991B
- Authority
- CN
- China
- Prior art keywords
- frame
- key frame
- block
- interpolated
- key
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000012545 processing Methods 0.000 claims abstract description 5
- 238000001914 filtration Methods 0.000 claims description 13
- 238000013441 quality evaluation Methods 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 10
- 238000012887 quadratic function Methods 0.000 claims description 9
- 238000001228 spectrum Methods 0.000 claims description 6
- 230000000007 visual effect Effects 0.000 claims description 5
- 230000009466 transformation Effects 0.000 claims description 2
- 230000011218 segmentation Effects 0.000 claims 2
- 238000010586 diagram Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 230000001133 acceleration Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000000903 blocking effect Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000006073 displacement reaction Methods 0.000 description 2
- 238000009499 grossing Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
Landscapes
- Television Systems (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention belongs to technical field of image processing, disclose a kind of side information generation method and device, to improve the accuracy of the side information of generation.A kind of side information generation method, including：Each key frame that time domain is adjoined in the video image of coding side output is obtained, wherein, the key frame is to be obtained in described image based on human eye feature, has interpolated frame among described two adjacent key frames；Obtain the forward motion vector of adjacent two key frame in the key frame；For any interpolated frame, motion compensation is carried out according to the information of the interpolated frame and the forward motion vector of the related keyword frame of the interpolated frame, generates side information.
Description
Technical field
The present invention relates to image processing techniques, more particularly to a kind of side information generation method and device.
Background technology
Distributed video coding frame is a kind of completely new asymmetrical video compress frame, and the system is in coding side to each
Video frame carries out absolute coding, i.e. intraframe coding carries out combined decoding, i.e. decoding inter frames in decoding end to each video frame.
In distributed video coding system, the information obtained by damaging message sink coding (Wyner-Ziv codings) is referred to as
For main information, side information (Side information) is the estimation for main information generated in decoding end.Side information
To decoder to help in the decoding process of main information, auxiliary decoder device is decoded.
System will obtain excellent distortion performance and compression efficiency, and most efficient method is exactly to use up the side information of generation
Amount is close to the current actual value for being intended to decoding frame, and therefore, the generation of side information is a research emphasis of distributed video coding.But
It is that inventor has found in the implementation of the present invention, the side information obtained by general side information generation method can make
There is deviation in the motion vector for obtaining many blocks in obtained side information frame.Thus, the accuracy of side information is not high.
Invention content
The technical problem to be solved in the present invention is to provide a kind of side information generation method and devices, are believed with improving the side of generation
The accuracy of breath.
A kind of side information generation method, including：
Each key frame that time domain is adjoined in the video image of coding side output is obtained, wherein, the key frame is described
It is obtained in image based on human eye feature, there is interpolated frame among described two adjacent key frames；
Obtain the forward motion vector of adjacent two key frame in the key frame；
For any interpolated frame, transported according to the forward direction of the information of the interpolated frame and the related keyword frame of the interpolated frame
Dynamic vector carries out motion compensation, generates side information.
Wherein, the forward motion vector for obtaining adjacent two key frame in the key frame includes：
Adjacent two key frame is divided into the block of predefined size respectively；
Calculate absolute difference in adjacent two key frame between corresponding blocks and；
If absolute difference between the corresponding blocks and less than the first preset value, the forward direction in adjacent two key frame is closed
Block of pixels in key frame alternately block of pixels, for substituting the respective pixel block of the interpolated frame among two key frame of phase；
If absolute difference between the corresponding blocks and more than the second preset value, motion compensation is carried out to the corresponding blocks；
If absolute difference between corresponding blocks and more than first preset value and less than the second preset value, the correspondence is kept
Block.
Wherein, the forward motion vector of the related keyword frame of the information and interpolated frame according to the interpolated frame
Motion compensation is carried out, generation side information includes：
The movement locus of object decomposed on horizontal and vertical direction, utilizes two key frames before the interpolated frame
Between forward motion vector and the interpolated frame rear two crucial interframe forward motion vector, using quadratic function to object
Movement locus simulated；
According to analog result, preceding key frame and rear key of the match block of the interpolated frame relative to the interpolated frame are calculated
The distance of frame, and according to the distance adjust the preceding key frame and it is described after key frame with the weights into interpolated frame, generation
Side information.
Wherein, after each key frame that time domain is adjoined in the video image of the acquisition coding side output, the method is also
Including：
Low-pass filtering is carried out to each key frame.
Wherein, the method further includes：
Image quality evaluation is carried out to the side information of generation.
A kind of side information generating means, including：
Frame acquiring unit, for obtaining each key frame that time domain is adjoined in the video image of coding side output, wherein, it is described
Key frame is to be obtained in described image based on human eye feature, has interpolated frame among described two adjacent key frames；
Vector acquiring unit, for obtaining the forward motion vector of adjacent two key frame in the key frame；
Generation unit, for for any interpolated frame, according to the information of the interpolated frame and the correlation of the interpolated frame
The forward motion vector of key frame carries out motion compensation, generates side information.
Wherein, the vector acquiring unit includes：
Block division module, for adjacent two key frame to be divided into the block of predefined size respectively；
Computing module, for calculate absolute difference in adjacent two key frame between corresponding blocks and；
Processing module, if absolute difference between the corresponding blocks and less than the first preset value, by adjacent two key frame
In forward direction key frame in block of pixels alternately block of pixels, for substituting pair of the interpolated frame among two key frame of phase
Answer block of pixels；If absolute difference between the corresponding blocks and more than the second preset value, motion compensation is carried out to the corresponding blocks；If
Absolute difference between corresponding blocks and more than first preset value and less than the second preset value, keeps the corresponding blocks.
Wherein, the generation unit includes：
Analog module for decomposed on horizontal and vertical direction the movement locus of object, is inserted using described
The forward motion vector of the forward motion vector of two crucial interframe and rear two crucial interframe of the interpolated frame before value frame, uses
Quadratic function simulates the movement locus of object；
Generation module, before according to analog result, calculating the match block of the interpolated frame relative to the interpolated frame
The distance of key frame and rear key frame, and according to the distance adjust the preceding key frame and it is described after key frame with into interpolation
The weights of frame generate side information.
Wherein, the frame acquiring unit is additionally operable to carry out low-pass filtering to the video image, obtains each key frame.
Wherein, described device further includes：
Quality evaluation unit carries out image quality evaluation to the side information of generation.
The Advantageous achievement that the present invention obtains is：
The present invention obtains each key frame that time domain is adjoined in the video image that coding side exports first, wherein, the key
Frame is to be obtained in described image based on human eye feature, has interpolated frame among described two adjacent key frames, then obtains
The forward motion vector of adjacent two key frame in the key frame, and for any interpolated frame, according to the information of the interpolated frame
And the forward motion vector of the related keyword frame of the interpolated frame carries out motion compensation, and then generates side information.Due to this hair
It is bright middle using the key frame obtained based on human eye feature, according to the information of the interpolated frame and the related keyword of the interpolated frame
The forward motion vector of frame carries out motion compensation, and then generates side information, therefore, is obtained using methods and apparatus of the present invention
Side information accuracy it is higher.
Description of the drawings
Fig. 1 is the flow chart of the side information generation method of the present invention；
Fig. 2 is motion vector selection figure in this method；
Fig. 3 is smothing filtering schematic diagram in space in this method；
Fig. 4 is overlapped block motion compensation schematic diagram in this method；
Fig. 5 is inventive algorithm and traditional algorithm contrast schematic diagram in bi-directional motion estimation；
Fig. 6 is structure similitude method in this method (SSIM) method block diagram；
Fig. 7 is the schematic diagram of the side information generating means of the present invention；
Fig. 8 is the structure chart of the side information generating means of the present invention.
Specific embodiment
To make the technical problem to be solved in the present invention, technical solution and advantage clearer, below in conjunction with attached drawing and tool
Body embodiment is described in detail.
In distributed video coding, video sequence X is by odd-numbered frame X2k±1(k is integer) and even frame X2k(k is integer)
Composition.In the present invention, it is assumed that odd-numbered frame is key frame, even frame is WZ frames (interpolated frame).It equally may be assumed that even frame to close
Key frame, odd-numbered frame are WZ frames, only need to ensure that there are key frames before and after WZ frames.Then side information is mainly by adjoining time domain
Adjacent key frame X2k±1Motion compensated interpolation is carried out to generate.
As shown in Figure 1, the side information generation method of the present invention includes：
Step 11 obtains each key frame that time domain is adjoined in the video image that coding side exports, wherein, the key frame is
It is obtained in described image based on human eye feature, there is interpolated frame among described two adjacent key frames.
It in this step, can also be to described after obtaining each key frame that time domain is adjoined in the video image of coding side output
Each key frame carries out low-pass filtering.
First, with reference to the visual characteristic HVS of human eye (Human Visual System, human visual system), input is closed
Key frame carries out low-pass filtering treatment.The vision of the mankind has unique light characteristic, spatial frequency characteristic and time frequency characteristic,
According to the spatial frequency characteristic of vision, the details of the higher ingredient representative image of vision intermediate frequency rate is smaller, and eye detection ability is got over
Low, wherein human eye is most sensitive to the spatial frequency component of 3~6CPD).Wherein, low-pass filtering can remove figure with smoothed image
As noise, enhance the reliability of motion vector, so as to get motion vector be more nearly true vector field.It is basic to realize process
It is as follows：First to original image and target image frequency spectrum through two-dimensional discrete fourier transform (2-DDFT), give respectively according to HVS
Subband is divided, and carries out two-dimentional Fourier inverse transformation (2-DIDFT) to the low frequency sub-band frequency spectrum of acquisition, establishes the figure of low frequency sub-band
Picture.
Step 12, the forward motion vector for obtaining adjacent two key frame in the key frame.
In this step, backward motion estimation is carried out by the image of the low frequency sub-band to being established in step 11.In this hair
Estimation is mainly carried out between front and rear key frame using block matching algorithm in bright embodiment.
As shown in Fig. 2, set search window as 8 × 8 rectangle, search range is 12 × 12 rectangle, step-length 1.Its
Middle search window is the base unit for performing estimation, and search range refers to carrying out in reference frame X2k-1 (k is integer)
The size in the region of search, within its scope search and X2k+1 (k is integer) frame on the most similar block of current block, step-length be
The distance between former and later two match blocks when being scanned in key frame.Wherein, step-length can reduce the computation complexity of frame
And provide the rough estimate of a true motion vector field.
Block matching method has many criterion, and the present invention uses SAD (Sum of absolute differences, absolute difference
Value and).Its basic principle is：For m × n block, using following formula (1) as motion vector distortion function
In formula, X2k-1(i, j) is the pixel value of macro block in previous reference frame, X2k+1(i+dx, j+dy) is in latter reference frame
The pixel value of respective macroblock, the motion vector of (dx, dy) between two macro blocks.That the corresponding block of minimum value for taking sad value is
The block of best match.Using obtained motion vector as the candidate motion vector of each non overlapping blocks in WZ frames.Wherein, each
Candidate motion vector can be met at a bit with WZ frames, and the nearest motion vector in block center of the intersection point from motion vector to be asked is selected as this most
The motion vector of good match block.
In the present invention, optimization is made to block matching algorithm.Front and rear two adjacent key frames are divided into predefined size first,
Such as 8 × 8 fritter, calculate the summation absolute error values (SAD) of corresponding blocks.It is meanwhile default according to practical experience setting first
Value T1 and the second preset value T2.Each group sad value is analyzed and determined.If SAD is less than threshold value T1, motion change degree is smaller, right
The influence of interpolation-movement estimation is little, to simplify decoding process, its variation is neglected, by the forward direction in adjacent two key frame
Block of pixels in key frame alternately block of pixels, the block of pixels in WZ frames are directly replaced by the respective pixel block in forward direction key frame
Generation.If SAD>T2, the motion change between pixels illustrated block is more violent, and the size of each block of pixels is further subdivided into 4 × 4 son
Block, then to its motion compensation.If sad value, between T1 and T2, block of pixels size in motion compensation remains unchanged.Herein
In step, during Block- matching according to corresponding blocks sad value size adjustment search block size, can substantially reduced ghost image show
As.
By the above-mentioned means, obtain the forward motion vector of adjacent two key frame.
Step 13, for any interpolated frame, according to the information of the interpolated frame and the related keyword frame of the interpolated frame
Forward motion vector carry out motion compensation, generate side information.
In this step, bi-directional motion estimation is mainly carried out, including：The movement locus of object is carried out horizontal and vertical
It is decomposed on direction, is closed using the forward motion vector of before the interpolated frame two crucial interframe and rear the two of the interpolated frame
The forward motion vector of key interframe simulates the movement locus of object using quadratic function.According to analog result, institute is calculated
The match block of interpolated frame is stated relative to the preceding key frame of the interpolated frame and the distance of rear key frame, and is adjusted according to the distance
The preceding key frame and it is described after key frame with the weights into interpolated frame, generate side information.
The above process described in detail below.
After final motion vector field is obtained, interpolated frame can be mended by the bidirectional-movement in standard video coder frame
It repays to obtain.Movement locus of object decompose on horizontal and vertical direction by the present invention, according to two frames before interpolated frame and
The motion vector twice of latter key interframe, the correction criterion moved using quadratic function as block, to match not advising for object
Then move.
Assuming that the forward motion vector of interpolated frame front cross frame is (x1,y1), two crucial interframe propulsions arrow before and after interpolated frame
It measures as (x2, y2), (x is the movement of horizontal direction, and y is the movement on vertical direction), crucial inter frame temporal is t, then by meter
It calculates, during this period of time the acceleration in object horizontal direction and vertical direction is respectively ax, ay：
Interpolated frame is respectively (s relative to the horizontal displacement of previous key frame and vertical displacementx, sy)：
The present invention utilizes the information of three frames before and after interpolated frame, and the movement locus of object is simulated using quadratic function,
The acceleration and curvilinear motion of object can be effectively matched.In addition, using weighted compensation interpositioning, quadratic function movement locus is utilized
The weight of key frame before and after calculating eliminates the correlation calculations of coding side, succinct effective.
In addition, in this step, after obtaining being intended to the bi-directional motion vector of decoding frame, if a certain piece of motion vector
Direction and surrounding block the direction of motion it is inconsistent when, the frame after motion compensation just will appear very strong blocking artifact.
At this moment can motion vector be adjusted by spatial domain smoothing technique.Neighborhood averaging is one kind in the smoothing technique of spatial domain, at this
In method, a neighborhood is taken to all pixels in the video frame that estimates one by one, is then made even to all pixels value in neighborhood
, the value of gained is used for replacing original pixel value.
Wherein, S represents the size of neighborhood；(m, n) is the pixel in neighborhood；M is the number of all pixels in neighborhood.
Different weights can also be set here to different pixels, and such as Fig. 3 can set the weights of pixel B5 to be adjusted
It is 70%, weights are set as 030% after other pixel value sum-average arithmetics in neighborhood, and the pixel value that mean value can be made later in this way is more
What is added is accurate, meets truth.
Mitigate the influence of blocking artifact using a kind of overlapped block motion compensation technology.It is assumed that want the 8 of interpolation
× 8 macro block is B, before overlapped block motion compensation is carried out, it first is expanded to 12 × 12 from 8 × 8, adjacent macro block
Overlapping region to occur, if its motion vector is MV4=(i4,j4), the motion vector of 8 × 8 pieces of excess-three is respectively MV1=
(i1,j1)、MV2=(i2,j2)、MV3=(i3,j3).For to insert the macro different zones in the block of 5 values, the parameter of reconstruction will basis
The number of overlapping block determines.
As shown in figure 5, for there is the pixel value in overlapping region 1 with 4 blocks, calculation formula is：
For there is the pixel value in overlapping region 2 with 2 blocks, calculation formula is：
For with the pixel value in other pieces of nonoverlapping regions 3, calculation formula is：
Due to using the key frame that is obtained based on human eye feature in the present invention, according to the information of the interpolated frame and described
The forward motion vector of the related keyword frame of interpolated frame carries out motion compensation, and then generates side information, therefore, utilizes the present invention's
The side information accuracy that method is obtained is higher.
In addition, on the basis of method shown in Fig. 1, the method may also include：Image is carried out to the side information of generation
Quality evaluation.Wherein, traditional mode is carried out using PSNR (Peak Signal to Noise Ratio, Y-PSNR)
Image quality evaluation, although this method have many advantages, such as it is simple, be easily achieved, it is not intended that the vision of the mankind and psychology are felt
By the subjective assessment with people disconnects.The present invention carries out quality evaluation using improved structural similarity algorithm to the WZ frames of generation,
Its process is as shown in Figure 6.
Finally, this three parts joins together to form whole structural similarity：
For the value of SSIM closer to 1, side information generation quality is higher.
As shown in fig. 7, the side information generating means of the present invention, including：
Frame acquiring unit 51, for obtaining each key frame that time domain is adjoined in the video image of coding side output, wherein, institute
It is to be obtained in described image based on human eye feature to state key frame, has interpolated frame among described two adjacent key frames；Arrow
Acquiring unit 52 is measured, for obtaining the forward motion vector of adjacent two key frame in the key frame；Generation unit 53, for pair
In any interpolated frame, carried out according to the information of the interpolated frame and the forward motion vector of the related keyword frame of the interpolated frame
Motion compensation generates side information.
Wherein, the frame acquiring unit 51 is specifically used for carrying out low-pass filtering to the video image, obtains each pass
Key frame.
Wherein, the vector acquiring unit 52 includes：
Block division module, for adjacent two key frame to be divided into the block of predefined size respectively；Computing module, for calculating
Absolute difference in adjacent two key frame between corresponding blocks and；Processing module, if absolute difference between the corresponding blocks and small
In the first preset value, by the alternately block of pixels of the block of pixels in the forward direction key frame in adjacent two key frame, for replacing
For the respective pixel block of the interpolated frame among two key frame of phase；If absolute difference between the corresponding blocks and pre- more than second
If value, motion compensation is carried out to the corresponding blocks；If absolute difference between corresponding blocks and more than first preset value and being less than
Second preset value keeps the corresponding blocks.
Wherein, the generation unit 53 includes：
Analog module for decomposed on horizontal and vertical direction the movement locus of object, is inserted using described
The forward motion vector of the forward motion vector of two crucial interframe and rear two crucial interframe of the interpolated frame before value frame, uses
Quadratic function simulates the movement locus of object；Generation module, for according to analog result, calculating of the interpolated frame
With block relative to the preceding key frame of the interpolated frame and the distance of rear key frame, and the preceding key frame is adjusted according to the distance
With it is described after key frame with the weights into interpolated frame, generate side information.
As shown in figure 8, described device may also include：
Quality evaluation unit 54 carries out image quality evaluation to the side information of generation.
The operation principle of the device of the invention can refer to the description of preceding method embodiment.
Due to using the key frame that is obtained based on human eye feature in the present invention, according to the information of the interpolated frame and described
The forward motion vector of the related keyword frame of interpolated frame carries out motion compensation, and then generates side information, therefore, utilizes the present invention's
The side information accuracy that device is obtained is higher.
The above is the preferred embodiment of the present invention, it is noted that for those skilled in the art
For, without departing from the principles of the present invention, several improvements and modifications can also be made, these improvements and modifications
It should be regarded as protection scope of the present invention.
Claims (4)
1. a kind of side information generation method, which is characterized in that including：
Each key frame that time domain is adjoined in the video image of coding side output is obtained, wherein, the key frame is in described image
In obtained based on human eye feature, with reference to the visual characteristic HVS of human eye, low-pass filtering treatment is carried out to input key frame, it is described
Low-pass filtering treatment realizes that process includes：To the original image through two-dimensional discrete fourier transform (2-DDFT) and target image frequency
Spectrum gives subband segmentation, and two-dimentional Fourier inverse transformation (2- is carried out to the low frequency sub-band frequency spectrum of acquisition respectively according to HVS
DIDFT), the image of low frequency sub-band is established；There is interpolated frame among described two adjacent key frames；
Obtain the forward motion vector of adjacent two key frame in the key frame；
For any interpolated frame, sweared according to the propulsion of the information of the interpolated frame and the related keyword frame of the interpolated frame
Amount carries out motion compensation, generates side information；
The forward motion vector for obtaining adjacent two key frame in the key frame includes：
Adjacent two key frame is divided into the block of predefined size respectively；
Calculate absolute difference in adjacent two key frame between corresponding blocks and；
If absolute difference between the corresponding blocks and less than the first preset value, by the forward direction key frame in adjacent two key frame
In block of pixels alternately block of pixels, for substituting the respective pixel block of the interpolated frame among two key frame of phase；
If absolute difference between the corresponding blocks and more than the second preset value, each block of pixels is divided into sub-block, then to described right
Block is answered to carry out motion compensation；
If absolute difference between corresponding blocks and more than first preset value and less than the second preset value, the corresponding blocks are kept；
In the process, the size during Block- matching according to the size adjustment search block of the sad value of corresponding blocks；
It is described for any interpolated frame, transported according to the forward direction of the information of the interpolated frame and the related keyword frame of the interpolated frame
Dynamic vector carries out motion compensation, and generation side information includes：
The movement locus of object decomposed on horizontal and vertical direction, utilizes before the interpolated frame two crucial interframe
The forward motion vector of forward motion vector and front and rear two crucial interframe of the utilization interpolated frame, using quadratic function to object
The movement locus of body is simulated；
According to analog result, the preceding key frame and rear key frame of the match block of the interpolated frame relative to the interpolated frame are calculated
Distance, and according to the distance adjust the preceding key frame and it is described after key frame with the weights into interpolated frame, generation side letter
Breath；
Image quality evaluation is carried out to the side information of generation, is calculated between the image and original image that decode and obtain by the following formula
Pixel deviations：
Wherein,
In formula, fk (x, y) and fk-1 (x, y) are respectively the gray value of pixel in primitive frame and decoding frame, and M × N is image
Size.
2. according to the method described in claim 1, it is characterized in that, it is described acquisition coding side output video image in time domain
After each key frame adjoined, the method further includes：
Low-pass filtering is carried out to each crucial frame video image.
3. a kind of side information generating means, which is characterized in that including：
Frame acquiring unit, for obtaining each key frame that time domain is adjoined in the video image of coding side output, wherein, the key
Frame is to be obtained in described image based on human eye feature, and with reference to the visual characteristic HVS of human eye, low pass is carried out to input key frame
Filtering process, the low-pass filtering treatment realize that process includes：To the original image through two-dimensional discrete fourier transform (2-DDFT)
With target image frequency spectrum, give subband segmentation respectively according to HVS, and it is anti-to carry out two-dimentional Fourier to the low frequency sub-band frequency spectrum of acquisition
It converts (2-DIDFT), establishes the image of low frequency sub-band；There is interpolated frame among described two adjacent key frames；
Vector acquiring unit, for obtaining the forward motion vector of adjacent two key frame in the key frame；
Generation unit, for for any interpolated frame, according to the information of the interpolated frame and the related keyword of the interpolated frame
The forward motion vector of frame carries out motion compensation, generates side information；
The vector acquiring unit includes：
Block division module, for adjacent two key frame to be divided into the block of predefined size respectively；
Computing module, for calculate absolute difference in adjacent two key frame between corresponding blocks and；
Processing module, will be in adjacent two key frame if absolute difference between the corresponding blocks and less than the first preset value
Block of pixels in forward direction key frame alternately block of pixels, for substituting the correspondence picture of the interpolated frame among two key frame of phase
Plain block；If absolute difference between the corresponding blocks and more than the second preset value, each block of pixels is divided into sub-block, then to described right
Block is answered to carry out motion compensation；If absolute difference between corresponding blocks and more than first preset value and less than the second preset value, protect
Hold the corresponding blocks；In the process, according to the big of the size adjustment search block of the sad value of corresponding blocks during Block- matching
It is small；
The generation unit includes：
Analog module for decomposed on horizontal and vertical direction the movement locus of object, utilizes the interpolated frame
The forward motion vector of the forward motion vector of preceding two crucial interframe and front and rear two crucial interframe of the utilization interpolated frame, is adopted
The movement locus of object is simulated with quadratic function；
Generation module, for according to analog result, calculating the match block of the interpolated frame relative to the preceding key of the interpolated frame
The distance of frame and rear key frame, and according to the distance adjust the preceding key frame and it is described after key frame with into interpolated frame
Weights generate side information；
Quality evaluation unit carries out image quality evaluation to the side information of generation, and the quality evaluation unit passes through the following formula
Calculate the pixel deviations between image and original image that decoding obtains：
Wherein,
In formula, fk (x, y) and fk-1 (x, y) are respectively the gray value of pixel in primitive frame and decoding frame, and M × N is image
Size.
4. device according to claim 3, which is characterized in that the frame acquiring unit be additionally operable to the video image into
Row low-pass filtering obtains each key frame.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410536223.XA CN104243991B (en) | 2014-10-11 | 2014-10-11 | A kind of side information generation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410536223.XA CN104243991B (en) | 2014-10-11 | 2014-10-11 | A kind of side information generation method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104243991A CN104243991A (en) | 2014-12-24 |
| CN104243991B true CN104243991B (en) | 2018-06-12 |
Family
ID=52231211
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410536223.XA Active CN104243991B (en) | 2014-10-11 | 2014-10-11 | A kind of side information generation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104243991B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106331721B (en) * | 2015-06-15 | 2019-06-28 | 中国科学院深圳先进技术研究院 | True motion estimation method |
| CN105939475A (en) * | 2016-06-06 | 2016-09-14 | 中国矿业大学 | High quality side information production method |
| CN115297313B (en) * | 2022-10-09 | 2023-04-25 | 南京芯视元电子有限公司 | Micro display dynamic compensation method and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7627037B2 (en) * | 2004-02-27 | 2009-12-01 | Microsoft Corporation | Barbell lifting for multi-layer wavelet coding |
| CN101835044A (en) * | 2010-04-23 | 2010-09-15 | 南京邮电大学 | Grouping method in frequency domain distributed video coding |
| CN103475879A (en) * | 2013-09-10 | 2013-12-25 | 南京邮电大学 | Side information generation method in distribution type video encoding |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7916791B2 (en) * | 2006-06-16 | 2011-03-29 | International Business Machines Corporation | Method and system for non-linear motion estimation |
-
2014
- 2014-10-11 CN CN201410536223.XA patent/CN104243991B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7627037B2 (en) * | 2004-02-27 | 2009-12-01 | Microsoft Corporation | Barbell lifting for multi-layer wavelet coding |
| CN101835044A (en) * | 2010-04-23 | 2010-09-15 | 南京邮电大学 | Grouping method in frequency domain distributed video coding |
| CN103475879A (en) * | 2013-09-10 | 2013-12-25 | 南京邮电大学 | Side information generation method in distribution type video encoding |
Non-Patent Citations (1)
| Title |
|---|
| 基于分层变块大小运动估计的边信息提取算法(英文);刘荣科等;《Chinese Journal of Aeronautics》;20090415(第02期) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104243991A (en) | 2014-12-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US7860167B2 (en) | Apparatus and method for adaptive 3D artifact reducing for encoded image signal | |
| Wang et al. | Frame rate up-conversion using trilateral filtering | |
| Kang et al. | Motion compensated frame rate up-conversion using extended bilateral motion estimation | |
| Zhai et al. | A low complexity motion compensated frame interpolation method | |
| Kang et al. | Dual motion estimation for frame rate up-conversion | |
| CN110741640B (en) | Optical flow estimation for motion compensated prediction in video coding | |
| TWI495328B (en) | Methods and apparatus for adaptively choosing a search range for motion estimation | |
| JP5203379B2 (en) | Spatial convention guidance time prediction for video compression | |
| CN105939475A (en) | High quality side information production method | |
| JP2008508751A5 (en) | ||
| US20110206127A1 (en) | Method and Apparatus of Frame Interpolation | |
| KR20120003147A (en) | Depth map coding and decoding apparatus using loop-filter | |
| US20150245071A1 (en) | Motion-compensated temporal filtering based on variable filter parameters | |
| TW200934222A (en) | Method and integrated circuit for video processing | |
| CN104602028A (en) | Entire frame loss error concealment method for B frame of stereoscopic video | |
| Aziz et al. | Motion estimation and motion compensated video compression using DCT and DWT | |
| CN104243991B (en) | A kind of side information generation method and device | |
| CN103475879A (en) | Side information generation method in distribution type video encoding | |
| Choi et al. | Triple-frame-based bi-directional motion estimation for motion-compensated frame interpolation | |
| CN103152566B (en) | A kind of video frame rate method for improving | |
| CN117750020A (en) | Method, system, equipment and storage medium for learning video coding | |
| KR100254667B1 (en) | Method and apparatus for detecting occulusion | |
| CN101600105A (en) | Frame frequency lifting method | |
| KR20110097868A (en) | Adaptive frequency domain filtering for phase plane correlation | |
| Agarwal | Compressing stereo images using a reference image and the exhaustive block matching algorithm to estimate disparity between the two images |
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 | ||
| GR01 | Patent grant |