CN106559670A - A kind of improved piecemeal video compress perception algorithm - Google Patents
A kind of improved piecemeal video compress perception algorithm Download PDFInfo
- Publication number
- CN106559670A CN106559670A CN201610976423.6A CN201610976423A CN106559670A CN 106559670 A CN106559670 A CN 106559670A CN 201610976423 A CN201610976423 A CN 201610976423A CN 106559670 A CN106559670 A CN 106559670A
- Authority
- CN
- China
- Prior art keywords
- fritter
- reference frame
- block
- sampling
- frame
- 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.)
- Pending
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/102—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
- H04N19/132—Sampling, masking or truncation of coding units, e.g. adaptive resampling, frame skipping, frame interpolation or high-frequency transform coefficient masking
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
- H04N19/17—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object
- H04N19/176—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object the region being a block, e.g. a macroblock
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/503—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction
- H04N19/51—Motion estimation or motion compensation
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The present invention relates to computer vision field, refers in particular to a kind of improved piecemeal video compress perception algorithm.Video acquisition and compressed encoding are combined while carry out, be, using the redundancy on video time axle, different sampling policies to be used to reference frame and non-reference frame, for reference frame, piecemeal is first carried out, be then fixed high sampling rate measurement；For non-reference frame, will make comparisons and then adjustment sampling policy with reference frame corresponding blocks after piecemeal.The sampling of non-reference frame can provide more information for reference frame so that higher video quality is obtained in the case where number of samples is little.Algorithm can be adaptively adjusted sampling rate according to the texture complexity degree inside frame of video simultaneously, optimize allocation of resources.Relative to general compression sampling algorithm, it is possible to reduce sampled value, the result for obtaining not only had met eye-observation but also had had high signal to noise ratio.
Description
Technical field
The present invention relates to computer vision field, refers in particular to a kind of improved piecemeal video compress perception algorithm.
Background technology
In visual communication, due to due to device memory is little and the network bandwidth is narrower, how data compression has been carried out
Become the key of video monitoring, in order to obtain high-quality video image, the approach for solving need to be sought from compression algorithm.
In recent years, compressed sensing has the advantages that prominent and wide application prospect in field of signal processing.Compressed sensing is calculated
Method has broken the principle that signal sampling speed must comply with nyquist sampling theorem, reduces signals collecting and hardware is wanted
Ask.On the other hand, compressed sensing framework allows to combine signals collecting and compression, directly obtains compressed signal, can
To come during the content of signal encryption is added to this.
From unlike image, video is a three dimensional signal, not only has openness in spatial domain, between frame of video
Similarity is also required to the design in view of algorithm.The substrate launched as signal using 3D wavelet transformations, during reconstruct once
All frame of video are reconstructed, the computation complexity that this algorithm needs is too high.Jing Zheng etc. notice frame of video it
Between similarity, the compressed sensing of routine has been carried out to reference frame, for non-reference frame is only between reference frame and non-reference frame
Residual error sampled, be superimposed during reconstruct again and obtained final result.But as non-reference frame depends on reference frame
Quality of recovery, if the Quality of recovery of reference frame is poor, video quality drastically will decline.Z.Liu etc. proposes frame of video and adopts
During sample, piecemeal is processed, and is then judged difference between fritter using related blocks, then using different sampling policies, is effectively improved
Efficiency of algorithm, but subsequent frames rely on the reconstruction quality of previous frame all the time, and the problem for having error accumulation occurs.
For above-mentioned problem, the present invention proposes a kind of new improved piecemeal video compress perception algorithm, is dividing
Under the framework of block sampling, sampling policy is adjusted according to the difference of related blocks, and with the sampling compensation previous frame of subsequent frames, is made
The quality of subsequent frames not exclusively relies on previous frame, moreover it is possible to improve the quality of previous frame, is improved whole video quality.
The content of the invention
The technical problem to be solved in the present invention is：For this particular problem of video compress, in order to improve compression backsight
Frequency result had not only met eye-observation but also had had high signal to noise ratio, it is proposed that a kind of improved piecemeal video compress perception algorithm.
The technical scheme of the algorithm specifically includes following steps：
(1) the fixed high sampling rate measurement of reference frame；
(2) the variable sampling rate measurement of non-reference frame；
(3) little block sort, sampling and reconstruct.
As the further improvement of technical solution of the present invention, the step 1) in, using block sampling configuration, by reference frame not
Overlap partition obtains the image block of K B × B size, and each image block identical M × N-dimensional calculation matrix Φ is individually adopted
Sample；
IfFor k-th fritter in t frame, be stretched as a vectorial form, block size be n ×
N, claimsForRelated blocks, they be located at different frame same positions；
For i-th (i=1,2 ..., K) individual image blockIts measurement vector can be expressed as：
In formulaRM×NIt is M*N dimension calculation matrix,It is then whole
The sampling process of reference frame can be expressed as Yt-1=Φ Xt-1, wherein Xt-1∈RN×KAnd Yt-1∈RM×KColumn vector be respectively it is each
The pixel vectors and measurement vector of image block.
As the further improvement of technical solution of the present invention, in the step (2), first by identical bits in fritter and reference frame
The fritter put is compared, and is then sampled according to result adjustment, specific as follows：
IfFor k-th fritter in t frame, be stretched as a vectorial form, block size be n ×
N, claimsForRelated blocks, they be located at different frame same positions；Using the difference of related blocks as judging fritter
The foundation of type, according toL1Norm andVariance adjustment sampling.
As the further improvement of technical solution of the present invention, the step 3) include：
When sampling to reference frame, frame of video is divided into into the fritter that nonoverlapping size is n × n first, it is then right
Each fritter carries out normal sample, i.e.,
When sampling to non-reference frame, frame of video is divided into into the fritter that nonoverlapping size is n × n first equally, then
Collection fritter one piece of measured value of very little,Wherein ΦpFor the calculation matrix of non-reference frame,For Φ0's
A part, Mp＜ ＜ M0；Then with fritter in reference frameMeasurement be compared,Calculate yrL1
Norm and variance the foundation in this, as little block sort；
L=σ (yr)+λ||yr||1 (2)
In formula, λ is a normaliztion constant, for adjusting variances sigma (yr) and l1Proportion between norm two indices；
M0For fixed high sampling, M1For extra samples, M2Sample for frame difference, M3Sample for conventional compact, according to L's
Fritter is divided into following 3 class by value：
(1) if L is ＜ T1, T1For the threshold value that the image block of standard video sequence sets, the spatial domain change between two fritters
It is small enough to and ignores, by fritterIt is labeled as static block；
(2) if T1≤ L ＜ T2, T2For the threshold value that the image block of standard video sequence sets, T1＜ T2, then by the fritter quilt
Less piece of change is labeled as, the difference of fritter is measured,SettingM2＜ M0；
In reconstruction stage, fritter residual error is first recovered, is then addedFinally give reconstruction result；IfTo become
Change smaller fritter, then no matterWhy type, by current blockChange is adjusted to than larger fritter, to be prevented effectively from
The accumulation of reconstruction error；
(3) if L >=T2, willChange is labeled as than larger block or dynamic block, and fritter is normally measured；
In formula：M3＞ ＞ M0；
In reconstruction stage, first look at the number of measured value to differentiate the type of current fritter, normal blocks and change are compared
Big blockMeasured using normal compressed sensing algorithm, but assumedIt is not enough to recover former fritter well, then examines
Look into Fritter isRelated blocks, ifContaining not motion block in fritter, motion block is not containing textured
Consistent fritter, then related blocksMeasured value can be byIt is used, by two frittersWithRecover together, i.e.,
For less piece of change, the fritter of non-reference frame is due to the residual error measured value for there was only fritterIt is available, the reconstruction result of reference frame is depended on, i.e.,
Compared with prior art, the invention has the advantages that：
Video acquisition and compressed encoding are combined while carry out, be using the redundancy on video time axle, it is right
Reference frame and non-reference frame use different sampling policies, for reference frame, first carry out piecemeal, are then fixed high sampling rate
Measurement；For non-reference frame, will make comparisons and then adjustment sampling policy with reference frame corresponding blocks after piecemeal.The sampling of non-reference frame
More information can be provided for reference frame so that higher video quality is obtained in the case where number of samples is little.Simultaneously
Algorithm can be adaptively adjusted sampling rate according to the texture complexity degree inside frame of video, optimize allocation of resources.Relative to
General compression sampling algorithm, it is possible to reduce sampled value, the result for obtaining not only had met eye-observation but also had had high signal to noise ratio.
Description of the drawings
Fig. 1 is improved piecemeal video compress perception algorithm flow chart described in embodiment.
Fig. 2 is each method described in embodiment and improved piecemeal video compress perception algorithm effect contrast figure.
Specific embodiment
Now by taking common single-frame imagess as an example, the present invention is described in further details with reference to accompanying drawing.
The present invention proposes a kind of improved piecemeal video compress perception algorithm, and the algorithm is comprised the following steps that：
The first step：The fixed high sampling rate measurement of reference frame：
For the single-frame imagess of n × n, if as a n2× 1 vector is directly sampled, then need m × n2Measurement
To obtain m measured value, n is the side dimensions length of single-frame imagess to matrix, and m is the number of calculation matrix, in this case, is surveyed
The storage and calculating of moment matrix is all very big, therefore the present invention adopts block sampling configuration, and by reference frame, overlap partition does not obtain K B
The image block of × B sizes, and each image block identical M × N-dimensional calculation matrix Φ is individually sampled.IfFor t when
K-th fritter in frame is carved, a vectorial form is stretched as, block size is n × n, is claimedForCorrelation
Block, they are located at the same position of different frame.For i-th (i=1,2 ..., K) individual image blockIts measurement vector can be with table
It is shown as：
In formulaRM×NIt is M*N dimension calculation matrix, K is the total of image block
Number, M are to take than (mN)/n2Little maximum integer,Then the sampling process of whole reference frame can be expressed as Yt-1=
ΦXt-1, wherein Xt-1∈RN×KAnd Yt-1∈RM×KColumn vector be respectively each image block pixel vectors and measurement vector.
Second step：The variable sampling rate measurement of non-reference frame；
For non-reference frame, first the fritter of same position in fritter and reference frame is compared, is then adjusted according to result
Whole sampling, it is specific as follows：
IfFor k-th fritter in t frame, be stretched as a vectorial form, block size be n ×
N, claimsForRelated blocks, they be located at different frame same positions.It is due to the similarity between consecutive frame, related
The difference of block can reflect the relation between related blocks, using this difference as the foundation for judging little block type.But due to
It is unknown in video acquisition stage fritter, the measured value of fritter can only be obtained.The distance of measurement space can be reacted well
Go out the distance of primary signal, therefore the difference of fritter measured valueCorrelation is reflected to a certain extent can
Relation between block.Two kinds of indexs are present invention employs, one kind is yrL1Norm, another kind are yrVariance.l1Norm can be with
Embody yrIntensity, i.e., the amount of pixel change between two fritters, and variance can embody the openness of fritter residual error, according to upper
State two kinds of index adjustment samplings.
3rd step：Little block sort, sampling and reconstruct；
When sampling to reference frame, due to no prior information, frame of video is divided into into nonoverlapping size for n first
The fritter of × n, then carries out normal sample to each fritter, i.e.,Wherein Φ0For the calculation matrix of reference frame.
When sampling to non-reference frame, frame of video is divided into into the fritter that nonoverlapping size is n × n first equally, then
The measured value of one piece of collection fritter very little,Wherein ΦpFor the calculation matrix of non-reference frame,For Φ0
A part, Mp＜ ＜ M0.Then with fritter in reference frameMeasurement be compared,Calculate yr's
l1Norm and variances sigma (yr) and in this, as the foundation of little block sort.
L=σ (yr)+λ||yr||1 (2)
In formula, λ is a normaliztion constant, for adjusting variances sigma (yr) and l1Norm | | yr||1Ratio between two indices
Weight.
M0For fixed high sampling, M1For extra samples, M2Sample for frame difference, M3Sample for conventional compact.According to L's
Fritter is divided into following 3 class by value：
(1) if L is ＜ T1, T1For the threshold value that the image block of standard video sequence sets, the spatial domain change between two fritters
It is small enough to and ignores, then claims fritterFor static block, due to the texture of the two blocks it is almost identical, it is not necessary to again to this
Fritter is measured, and can directly use fritterMeasured value.But ifMeasurement be not enough to reconstruct it is satisfactory
Result, the reconstruction quality of the two fritters will all be affected, so extra measurement is necessary, this fritter will be entered
The extra measurement of row, to guarantee the Quality of recovery in the case where limited quantity is measured, the measured value of highly similar fritter mutually can be handed over
Change, while being recovered.
(2) if T1≤ L ＜ T2, T2For the threshold value that the image block of standard video sequence sets, T1＜ T2, then the block will be by
It is labeled as less piece of change.Because the measurement difference of two fritters is less, between two pieces, still there are very strong dependency, explanation
Residual error between the two fritters will be a very sparse signal, and openness than fritter itself is eager to excel a lot.Therefore will
The difference of fritter is measured,Due to fritter difference it is openness very strong, it is only necessary to little survey
Value can recover satisfied result, therefore setM2＜ M0。
In reconstruction stage, fritter residual error can be first recovered, is then addedFinally give reconstruction result.Current blockReconstruction quality will depend onQuality, ifDo not rebuild well,The error that reconstruction is caused can be tired out
Product is to current blockThe fritter of next frame can even be accumulated.In order to avoid the accumulation of error, it is necessary to carry out type tune to fritter
It is whole.IfTo change smaller fritter, then no matterWhy type, by current blockIt is adjusted to and changes less than larger
Block.The accumulation of error can be so prevented effectively from.
(3) if L >=T2, this illustrates that the dependency of two fritters is very low, and texture has a very large change.WillLabelling
It is to change than larger block or dynamic block, in order to ensure reconstruction quality, fritter will be normally measured, but the number of measured value
Amount will be greatly enhanced,
In formula：M3＞ ＞ M0, compare difficult acquisition and compare M0Many measured values, actually (1)-two kinds of (2) class is little
The measured value of block distribution is all than the number M of normal measurement0It is many less, so M3M can be much larger than0, change than larger region
Quality of recovery can be guaranteed.
In measuring phases in order to obtain higher measurement efficiency, adaptive Measurement Algorithm is devised, under different situations
Fritter use different Measurement Algorithm, so corresponding restructing algorithm must be designed to obtain best reconstruction result.First
Different types of piece all has been assigned different number of measured value, therefore in reconstruction stage, just first looks at the number of measured value
The type of current fritter can be differentiated, has the conventional bar of reference frame, the not motion block of non-reference frame changes little block and changes greatly
Block.
Normal blocks and change are than larger blockMeasured using normal compressed sensing algorithm, it is assumed thatNo
Be enough to recover former fritter well, checkThese fritters areRelated blocks, if these fritters
In containing not motion block, that is, containing textured consistent fritter, then the measured value of related blocks can be byIt is used, can be by
Two fritters recover together.Under this policy, fritter obtains more measured values, and Quality of recovery will be obtained
Must improve.
For less piece of change, the fritter of non-reference frame is due to the residual error measured value for there was only fritterIt is available, the reconstruction result of reference frame is depended on, i.e.,However it is possible to next frame this
Individual fritter can be changed into not motion block, so be obtained with more preferable effect.
The method proposed in the present invention can actually be embedded in FPGA realizations, and exploitation has the camera of video compression functionality or takes the photograph
Camera.Above example only plays a part of to explain technical solution of the present invention that protection domain of the presently claimed invention does not limit to
In realizing system and specific implementation step described in above-described embodiment.Therefore, only to specific formula and calculation in above-described embodiment
Method is simply replaced, but still consistent with the method for the invention technical scheme of its flesh and blood, all should belong to the present invention
Protection domain.
Claims (4)
1. a kind of improved piecemeal video compress perception algorithm, is characterized in that, comprise the steps：
1) the fixed high sampling rate measurement of reference frame；
2) the variable sampling rate measurement of non-reference frame；
3) little block sort, sampling and reconstruct.
2. a kind of improved piecemeal video compress perception algorithm according to claim 1, it is characterised in that the step 1)
In, using block sampling configuration, by reference frame, overlap partition does not obtain the image block of K B × B size, and each image block is used
Identical M × N-dimensional calculation matrix Φ individually samples；
IfFor k-th fritter in t frame, a vectorial form is stretched as, block size is n × n, is claimedForRelated blocks, they be located at different frame same positions；
For i-th (i=1,2 ..., K) individual image blockIts measurement vector can be expressed as：
In formulaRM×NIt is M*N dimension calculation matrix,It is then whole to refer to
The sampling process of frame can be expressed as Yt-1=Φ Xt-1, wherein Xt-1∈RN×KAnd Yt-1∈RM×KColumn vector be respectively each image
The pixel vectors and measurement vector of block.
3. a kind of improved piecemeal video compress perception algorithm according to claim 1, is characterized in that, the step (2)
In, first the fritter of same position in fritter and reference frame is compared, is then sampled according to result adjustment, it is specific as follows：
IfFor k-th fritter in t frame, a vectorial form is stretched as, block size is n × n, is claimedForRelated blocks, they be located at different frame same positions；Using the difference of related blocks as judging little block type
Foundation, according toL1Norm andVariance adjustment sampling.
4. a kind of improved piecemeal video compress perception algorithm according to Claims 2 or 3, is characterized in that, the step
3) include：
When sampling to reference frame, frame of video is divided into into the fritter that nonoverlapping size is n × n first, then to each
Fritter carries out normal sample, i.e.,
When sampling to non-reference frame, frame of video is divided into into the fritter that nonoverlapping size is n × n first equally, is then gathered
One piece of measured value of fritter very little,Wherein ΦpFor the calculation matrix of non-reference frame,For Φ0One
Point, Mp＜ ＜ M0；Then with fritter in reference frameMeasurement be compared,Calculate yrL1Norm
With variance and in this, as the foundation of little block sort；
L=σ (yr)+λ|yr||1 (2)
In formula, λ is a normaliztion constant, for adjusting variances sigma (yr) and l1Proportion between norm two indices；
M0For fixed high sampling, M1For extra samples, M2Sample for frame difference, M3Sample for conventional compact, will be little according to the value of L
Block is divided into following 3 class：
(1) if L is ＜ T1, T1For the threshold value that the image block of standard video sequence sets, the spatial domain change between two fritters is little to arrive
Be enough to ignore, by fritterIt is labeled as static block；
(2) if T1≤ L ＜ T2, T2For the threshold value that the image block of standard video sequence sets, T1＜ T2, then the fritter is labeled
For less piece of change, the difference of fritter is measured,SettingM2＜ M0；
In reconstruction stage, fritter residual error is first recovered, is then addedFinally give reconstruction result；IfFor change ratio
Less fritter, then no matterWhy type, by current blockChange is adjusted to than larger fritter, to be prevented effectively from reconstruction
The accumulation of error；
(3) if L >=T2, willChange is labeled as than larger block or dynamic block, and fritter is normally measured；
In formula：M3＞ ＞ M0；
In reconstruction stage, first look at the number of measured value to differentiate the type of current fritter, normal blocks and change are than larger
BlockMeasured using normal compressed sensing algorithm, but assumedIt is not enough to recover former fritter well, then checksFritter isRelated blocks, ifContaining not motion block in fritter, motion block is not containing textured one
The fritter of cause, then related blocksMeasured value can be byIt is used, by two frittersWithRecover together, i.e.,
For less piece of change, the fritter of non-reference frame is due to the residual error measured value for there was only fritterCan
With depending on the reconstruction result of reference frame, i.e.,
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610976423.6A CN106559670A (en) | 2016-11-07 | 2016-11-07 | A kind of improved piecemeal video compress perception algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610976423.6A CN106559670A (en) | 2016-11-07 | 2016-11-07 | A kind of improved piecemeal video compress perception algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106559670A true CN106559670A (en) | 2017-04-05 |
Family
ID=58443953
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610976423.6A Pending CN106559670A (en) | 2016-11-07 | 2016-11-07 | A kind of improved piecemeal video compress perception algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106559670A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107690070A (en) * | 2017-08-23 | 2018-02-13 | 南通河海大学海洋与近海工程研究院 | Distributed video compression perceptual system and method based on feedback-less Rate Control |
| CN109087332A (en) * | 2018-06-11 | 2018-12-25 | 西安电子科技大学 | One kind being based on the relevant occlusion detection method of piecemeal |
| CN109559270A (en) * | 2018-11-06 | 2019-04-02 | 华为技术有限公司 | A kind of image processing method and electronic equipment |
| CN110418137A (en) * | 2019-07-31 | 2019-11-05 | 东华大学 | A kind of rest block collection measured rate regulation method for intersecting subset guiding |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050036547A1 (en) * | 2003-08-13 | 2005-02-17 | Anthony Vetro | Resource-constrained sampling of multiple compressed videos |
| CN103559695A (en) * | 2013-09-29 | 2014-02-05 | 哈尔滨工程大学 | Adaptive block compressive sensing reconstruction method |
| CN105678699A (en) * | 2015-05-06 | 2016-06-15 | 西安电子科技大学 | Compressed sensing image reconstruction method based on measurement domain block significance detection |
-
2016
- 2016-11-07 CN CN201610976423.6A patent/CN106559670A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050036547A1 (en) * | 2003-08-13 | 2005-02-17 | Anthony Vetro | Resource-constrained sampling of multiple compressed videos |
| CN103559695A (en) * | 2013-09-29 | 2014-02-05 | 哈尔滨工程大学 | Adaptive block compressive sensing reconstruction method |
| CN105678699A (en) * | 2015-05-06 | 2016-06-15 | 西安电子科技大学 | Compressed sensing image reconstruction method based on measurement domain block significance detection |
Non-Patent Citations (1)
| Title |
|---|
| 李涛 等: "一种新的基于分块的视频压缩感知算法", 《北京理工大学学报》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107690070A (en) * | 2017-08-23 | 2018-02-13 | 南通河海大学海洋与近海工程研究院 | Distributed video compression perceptual system and method based on feedback-less Rate Control |
| CN107690070B (en) * | 2017-08-23 | 2019-11-26 | 南通河海大学海洋与近海工程研究院 | Based on distributed video compression perceptual system and method without feedback code rate control |
| CN109087332A (en) * | 2018-06-11 | 2018-12-25 | 西安电子科技大学 | One kind being based on the relevant occlusion detection method of piecemeal |
| CN109559270A (en) * | 2018-11-06 | 2019-04-02 | 华为技术有限公司 | A kind of image processing method and electronic equipment |
| WO2020093988A1 (en) * | 2018-11-06 | 2020-05-14 | 华为技术有限公司 | Image processing method and electronic device |
| CN109559270B (en) * | 2018-11-06 | 2021-12-24 | 华为技术有限公司 | Image processing method and electronic equipment |
| CN110418137A (en) * | 2019-07-31 | 2019-11-05 | 东华大学 | A kind of rest block collection measured rate regulation method for intersecting subset guiding |
| CN110418137B (en) * | 2019-07-31 | 2021-05-25 | 东华大学 | Cross subset guided residual block set measurement rate regulation and control method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105451019B (en) | A kind of image compression transmission method towards wireless video sensor network | |
| Wan et al. | Compressive image fusion | |
| US9159121B2 (en) | Method for image denoising | |
| CN106559670A (en) | A kind of improved piecemeal video compress perception algorithm | |
| CN104123705B (en) | A kind of super-resolution rebuilding picture quality Contourlet territory evaluation methodology | |
| CN110827198A (en) | Multi-camera panoramic image construction method based on compressed sensing and super-resolution reconstruction | |
| CN102034250A (en) | Edge structure information based block compression perception reconstruction method | |
| CN102123278A (en) | Signal source encoding method based on distributed compressive sensing technology | |
| Tang et al. | Bayesian framework with non-local and low-rank constraint for image reconstruction | |
| Nirmalraj et al. | Fusion of visible and infrared image via compressive sensing using convolutional sparse representation | |
| Cui et al. | Deep neural network based sparse measurement matrix for image compressed sensing | |
| Chen et al. | Robust Kronecker product video denoising based on fractional-order total variation model | |
| US20150169632A1 (en) | Method and apparatus for image processing and computer readable medium | |
| CN103596006A (en) | Image compression method based on vision redundancy measurement | |
| CN113284202B (en) | Image compression sensing method of scalable network based on content self-adaption | |
| CN114651270A (en) | Depth loop filtering by time-deformable convolution | |
| CN108573509B (en) | Self-adaptive compressed sensing image coding method based on measurement domain significance detection model | |
| CN107509074B (en) | Self-adaptive 3D video compression coding and decoding method based on compressed sensing | |
| Liu et al. | An end-to-end multi-scale residual reconstruction network for image compressive sensing | |
| CN108596851A (en) | A kind of compressed sensing image processing algorithm based on Wavelet transformation | |
| Xia et al. | A regularized tensor decomposition method with adaptive rank adjustment for compressed-sensed-domain background subtraction | |
| Li et al. | Image reconstruction for compressed sensing based on joint sparse bases and adaptive sampling | |
| Starck et al. | An overview of inverse problem regularization using sparsity | |
| Lakshman et al. | Image recovery using sparse reconstruction based texture refinement | |
| Lu et al. | A new non-local video denoising scheme using low-rank representation and total variation regularization |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170405 |