CN103517079B - Compression video acquisition based on data-driven subspace collection and reconfiguration system - Google Patents
Compression video acquisition based on data-driven subspace collection and reconfiguration system Download PDFInfo
- Publication number
- CN103517079B CN103517079B CN201310422841.7A CN201310422841A CN103517079B CN 103517079 B CN103517079 B CN 103517079B CN 201310422841 A CN201310422841 A CN 201310422841A CN 103517079 B CN103517079 B CN 103517079B
- Authority
- CN
- China
- Prior art keywords
- subspace
- collection
- block
- module
- subspace collection
- 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
Abstract
The invention provides a kind of compression video acquisition based on data-driven subspace collection and reconfiguration system, including: subspace collection constructing module, sparse basis array constructing module, video signal sensing module and reconstruction processing module, wherein: subspace collection constructing module utilizes clustering method generated subspace collection, sparse basis array constructing module utilizes the sparse base that linear subspaces learning method generated subspace collection is corresponding, video signal is projected by sensing module with the form of block, and the data of gained are finally decoded reconstruct in reconstruction processing module.The present invention has also agreed with the distributed gradual structure of video sampling process while providing compression sampling, special tectonic to sparse basis array also improves degree of accuracy and the efficiency of reconstruct, the present invention substantially increases the sampling efficiency of video signal, under different Sampling Compression rates, compare additive method achieve reconstruct gain, also possess good extensibility simultaneously.
Description
Technical field
The present invention relates to a kind of video signal and obtain scheme, a kind of compression video based on data-driven subspace collection
Gather and reconfiguration system.
Background technology
The collection of video signal and coding (compression) are applied most important for the storage of video and transmission etc..Traditional letter
Number processing system uses the pattern of recompression of first sampling: in order to intactly preserve all information of signal, should be with not less than signal
Video is sampled by the twice sample frequency of bandwidth；The primary signal collected reaches by after a series of coding techniques
Except the purpose of redundancy, the bottleneck of correlation technique is take substantial amounts of sensor and calculate resource just in order to obtain process
After a small amount of Signal Compression data, the resource requirement to sampling end is too high.Imitate to improve the collection of video signal further
Rate, adds some signal processing technologies while sampling, and one of which scheme is then sampling to be carried out with compression simultaneously,
Then by some algorithms of rear end, the data after compression are reconstructed.
Through finding the literature search of prior art, Ying Liu, Ming Li and Dimitris A.Pados is 2013
" IEEE Transactions on Circuits and Systems for Video Technology " (TCSVT) in year
" the Motion-Aware Decoding of Compressed-Sensed Video " literary composition delivered on periodical proposes
Based on Karhunen-Loeve transform(KLT) compression Application in Sensing come on video sampling by the reconstruct of base, and should
Video block is directly used sensing matrix to be compressed sampling at sample code end by method, uses KLT base conduct in decoding end
Signal is reconstructed by sparse base, and this method can be effectively improved the efficiency of video sampling, and ensures that reconstruct obtains
The subjective quality of video, but the KLT base that this method is used produces in Local Search window, multiple for having
Miscellaneous texture or the video scene of strenuous exercise, the KLT base that the method is used cannot be accurately and effectively to frame of video block
Carry out rarefaction representation, and then cause effect to reduce.It is a kind of more efficient that these deficiencies promote us to look on its basis
Ground reconstructing method, makes full use of the special construction of video signal block to improve the subjective and objective quality of reconstruction result.Yue M.Lu
With Minh N.Do on " IEEE Transactions on Signal Processing " (TSP) periodical in 2008
" the A Theory for Sampling Signals From a Union of Subspaces " literary composition delivered proposes
Signal sampling theory based on subspace collection, this theory gives the sampling of the signal for being in subspace collection to expire
The uniqueness of foot and the condition of stability, but the subspace collection that this theory is assumed is opened into by fixed base, it is impossible to carry
For significantly more efficient openness and adaptability.
Summary of the invention
The present invention is directed to the deficiencies in the prior art, it is provided that a kind of compression video acquisition based on data-driven subspace collection with
Reconfiguration system, can be effectively improved the subjective and objective quality of video signal collective efficiency and reconfiguration system, and can be as one
General video acquisition instrument.
The present invention is achieved by the following technical solutions:
Compression video acquisition based on data-driven subspace collection of the present invention and reconfiguration system, including: subspace collection
Constructing module, sparse basis array constructing module, video signal sensing module and reconstruction processing module, wherein:
Described subspace collection constructing module, to video signal key frame block, utilizes clustering method generated subspace collection, and should
Subspace collection exports the input of sparse basis array constructing module；
Described sparse basis array constructing module receives subspace collection, utilizes linear subspaces learning method generated subspace set pair
The sparse basis array answered, and this sparse basis array is exported the input of reconstruction processing module；
The non-key frame block of video signal is projected by described video signal sensing module with the form of block, is observed
Value, and this observation is exported the input of reconstruction processing module；
Described reconstruction processing module receives the basic matrix of described sparse basis array constructing module output and passes with described video signal
The measured value of sense module output, is reconstructed signal.
Described subspace collection constructing module, it is achieved generated different classes of block by carrying out block cluster on the key frame rebuild
Group.Every class block group is corresponding to a sub spaces, and the block group that cluster obtains is as sparse for generated subspace collection of training set
Basic matrix.Subspace based on block cluster structure can pass through sparse Subspace clustering method and block matching method to whole heavy
Structure key frame is operable to realize.
Described sparse basis array constructing module realizes a kind of orthonormal basis generated by linear subspaces learning method, line
Subspace learning method individually acts on different block groups and obtains different bases, and then composition sparse basis array.It
The immanent structure representing high dimensional signal of the property of can adapt to, can more effectively believe by rarefaction representation video relative to fixed base
Number, and the rarefaction representation that signal is in this sparse basis array is to have block structured.
Described sensing module is the digital micromirror device (DMD) of a kind of single order, and it simulates the compression to video signal and passes
Sense.
Described reconstruction processing module is by a kind of convex relaxed algorithm model realization.
The compression sensing technology integrated based on data-driven subspace used in the present invention provides as the collection of video signal
General solution, in particular for the video signal with complex texture and strenuous exercise.Son used in the present invention
Space collection is by using sparse subspace clustering and block matching method cluster to obtain in the key frame of reconstruct, fully profit
With the unique texture of frame of video block, and the space time redundancy between intra frame, improve efficiency and the performance of sampling.
On the other hand, the important function played in the restructuring procedure of compression sensing in view of sparse basis array, the present invention is by linear
The method of sub-space learning individually learns to obtain the sparse of corresponding base and then composition subspace collection to every sub spaces
Basic matrix, so enables to frame block signal and has adaptability rarefaction representation, and this rarefaction representation has structural, enters
And improve sampling efficiency (reducing the necessary hits needed for Accurate Reconstruction), moreover it is possible to accelerate the convergence of convex lax restructing algorithm
And stability, contribute to performance and the lifting of practicality of data-driven subspace of the present invention collection compression sensing.
Compared with prior art, the present invention has a following beneficial effect:
The present invention substantially increases reconstruction property, the video compress being reconstructed with traditional use fixed base or KLT base
Sensor-based system is compared, due to the reconstruct of the present invention use adaptive global optimum gene this on quality reconstruction all
Can be strengthened；For other high dimensional signal, the present invention is used as by suitable amendment, has stronger adaptation
Property；When rebuilding due to subspace collection and the special tectonic of sparse basis array so that signal has structural rarefaction representation,
Therefore the present invention can improve sampling efficiency in the case of the subjective effect not reducing video further, accelerates convex pine simultaneously
The convergence rate of relaxation restructing algorithm, compares additive method under different Sampling Compression rates and achieves reconstruct gain, the most also
Possesses good extensibility.
Accompanying drawing explanation
The detailed description made non-limiting example with reference to the following drawings by reading, other of the present invention is special
Levy, purpose and advantage will become more apparent upon:
Fig. 1 is the structured flowchart of present system one embodiment；
Fig. 2 is subspace collection constructing module fundamental diagram；
Fig. 3 is the structural rarefaction representation schematic diagram that frame of video block signal is produced by sparse basis array constructing module.
Detailed description of the invention
Below in conjunction with specific embodiment, the present invention is described in detail.Following example will assist in those skilled in the art
Member is further appreciated by the present invention, but limits the present invention the most in any form.It should be pointed out that, the common skill to this area
For art personnel, without departing from the inventive concept of the premise, it is also possible to make some deformation and improvement.These broadly fall into
Protection scope of the present invention.
As it is shown in figure 1, the structured flowchart of one embodiment of the invention, including: subspace collection constructing module, sparse group moment
Battle array constructing module, video signal sensing module and reconstruction processing module, wherein: subspace collection constructing module utilizes cluster side
Method generated subspace collection, sparse basis array constructing module utilizes corresponding dilute of linear subspaces learning method generated subspace collection
Dredging base, sensing module is compressed projection to video signal with the form of block, and the observation of gained is finally at reconstruction processing mould
Block is decoded reconstruct.In coding side, video signal sensing module carries out sampling and produces measured value video signal；?
In decoding end, sparse basis array constructing module produces basic matrix；Described sparse basis array constructing module output basic matrix
Reconstruction processing module is entered, signal in reconstruction processing module together with the measured value of described video signal sensing module output
It is reconstructed.
In the present embodiment, described subspace collection constructing module is as in figure 2 it is shown, do block and gather in the key frame that view picture is rebuild
Class, wherein: the set of blocks X={x in key frame1,x2,…,xK, utilize sparse Subspace clustering method or Block-matching side
Method is divided into t cluster X X1,X2,…,Xt, the block in each cluster is similar and belongs to a sub spaces.
X1,X2,…,XtCorresponding to t sub spaces S1,S2,…,St, then arbitrary frame of video block signal x broadly falls into subspace collection
U=∪Si。
In the present embodiment, described sparse basis array constructing module realizes a kind of mark generated by linear subspaces learning method
Almost-orthogonal basis, linear subspaces learning method (such as principal component analysis (PCA)) individually acts on different block groups
Xi, i=1 ..., t obtains different base Ψi, i=1 ..., t, and then composition sparse basis array Ψ*=[Ψ1,Ψ2,…,Ψt].Should
The immanent structure representing frame of video block signal of the sparse basis array property of can adapt to, can be the dilutest relative to fixed base
Relieving the exterior syndrome shows video signal, and the rarefaction representation c that signal is in this sparse basis array*It is that there is block structured, such as Fig. 3 institute
Show.
In the present embodiment, described video signal sensing module is the digital micro mirror projection equipment (DMD) of a kind of single order, it
Simulating the compression to video signal sensing y=Φ x, Φ is stochastical sampling matrix.First key frame block is carried out by this invention
Compression sampling, sample rate is 0.7, then non-key frame block signal is compressed sampling, sample rate be chosen at 0.1
Between 0.6, sampling based on frame of video block improves the speed of video sampling and reconstruct.
In the present embodiment, described reconstruction processing module is by a kind of convex relaxed algorithm model realization, particularly as follows: right
In key frame, find l1The c of Norm minimum makes y=Φ Ψ c, and obtain is a globally optimal solution, uses two-dimensional dct
Base Ψ is multiplied by this globally optimal solution and can be obtained by the key frame block signal of required reconstruct；For non-key frame, find
l2, IThe c of Norm minimum*Make y=Φ Ψ*c*, obtain is a globally optimal solution, uses Ψ*It is multiplied by this global optimum
Solution can be obtained by the non-key frame block signal of required reconstruct.Wherein, Φ is stochastical sampling matrix, l2, INorm is mixing
Norm,I is the subscript of block group in block structure, such as Fig. 3.
Implementation result
Being set to of key parameter in the present embodiment: experiment video sequence derives from Football_cif.yuv(352x288
The YUV file of 4:2:0 form), altogether take 250 frames.Every ten frames are a frame group, and choosing the first frame is key frame,
Rear nine frames are non-key frame, and the selection of dimension of block is 16 × 16 pixels.Owing to the gray-scale map of signal has concentrated most energy
Amount, test mainly completes on gray-scale map.It is of the present invention based on data-driven that the present embodiment compares employing
The method of the compressed sensing of space collection and Ying Liu et al. are at " Motion-Aware Decoding of
Compressed-Sensed Video " method in paper, and Yue M.Lu et al. is at " A Theory for Sampling
Signals From a Union of Subspaces " method in paper.Used by the present invention, sparse base have chosen PCA base,
The dimension of every sub spaces is 10, and the number of the subspace that cluster produces is 50.
First two method is compared therewith, and when compression ratio is 0.2, the present embodiment system obtains 9.2dB, 2.7dB respectively
Reconstruct gain；When compression ratio is 0.3, the present embodiment system obtains the reconstruct gain of 11.6dB, 2.8dB respectively；
When compression ratio is 0.4, the present embodiment system obtains the reconstruct gain of 11.4dB, 4.2dB respectively；It is 0.5 in compression ratio
Time, the present embodiment system obtains the reconstruct gain of 10.3dB, 6.3dB respectively；
Experiment shows, the present embodiment system reconstructing video sequence out is substantially better than other two kinds of methods on reconstruction quality
The video sequence obtained.
Above the specific embodiment of the present invention is described.It is to be appreciated that the invention is not limited in
Stating particular implementation, those skilled in the art can make various deformation or amendment within the scope of the claims,
This has no effect on the flesh and blood of the present invention.
Claims (6)
1. a compression video acquisition based on data-driven subspace collection and reconfiguration system, it is characterised in that including:
Subspace collection constructing module, sparse basis array constructing module, video signal sensing module and reconstruction processing module, wherein:
Described subspace collection constructing module, to video signal key frame block, utilizes clustering method generated subspace collection, and should
Subspace collection exports the input of sparse basis array constructing module；
Described sparse basis array constructing module receives subspace collection, utilizes linear subspaces learning method generated subspace set pair
The sparse basis array answered, and this sparse basis array is exported the input of reconstruction processing module；Described sparse basis array
Constructing module realizes a kind of orthonormal basis of being generated by linear subspaces learning method, its property of can adapt to represent height
The immanent structure of dimensional signal, relative to fixed base can more effectively rarefaction representation video signal, this in sparse basis array
Rarefaction representation be that there is block structured；
The non-key frame block of video signal is projected by described video signal sensing module with the form of block, is observed
Value, and this observation is exported the input of reconstruction processing module；
Described reconstruction processing module receives the basic matrix of described sparse basis array constructing module output and passes with described video signal
The observation of sense module output, is reconstructed signal.
Compression video acquisition based on data-driven subspace collection the most according to claim 1 and reconfiguration system, its
Feature is, described subspace collection constructing module, it is achieved generated different classes of by carrying out block cluster on the key frame of reconstruct
Block group, every class block group corresponds to a sub spaces, the block group that obtains of cluster as training set for generated subspace collection
Sparse basis array.
Compression video acquisition based on data-driven subspace collection the most according to claim 2 and reconfiguration system, its
Feature is, subspace based on block cluster structure can be by sparse Subspace clustering method and block matching method to whole reconstruct
Key frame is operable to realize.
Compression video acquisition based on data-driven subspace collection the most according to claim 1 and reconfiguration system, its
Feature is, described linear subspaces learning method individually acts on different block groups and obtains different bases, and then group
Become sparse basis array.
5. according to the compression video acquisition based on data-driven subspace collection described in any one of claim 1-3 and reconstruct
System, is characterized in that, described video signal sensing module is the digital micromirror device of a kind of single order, and it simulates regarding
Frequently the compression sensing of signal.
6. according to the compression video acquisition based on data-driven subspace collection described in any one of claim 1-3 and reconstruct
System, is characterized in that, described reconstruction processing module is by a kind of convex relaxed algorithm model realization, and the overall situation found is
Excellent solution is multiplied by the reconstruction signal that sparse base seeks to obtain.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310422841.7A CN103517079B (en) | 2013-09-16 | 2013-09-16 | Compression video acquisition based on data-driven subspace collection and reconfiguration system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310422841.7A CN103517079B (en) | 2013-09-16 | 2013-09-16 | Compression video acquisition based on data-driven subspace collection and reconfiguration system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103517079A CN103517079A (en) | 2014-01-15 |
| CN103517079B true CN103517079B (en) | 2016-11-16 |
Family
ID=49898992
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310422841.7A Active CN103517079B (en) | 2013-09-16 | 2013-09-16 | Compression video acquisition based on data-driven subspace collection and reconfiguration system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103517079B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104243986B (en) * | 2014-09-15 | 2018-07-20 | 上海交通大学 | Compression video acquisition and reconfiguration system based on data-driven tensor subspace |
| CN104301728B (en) * | 2014-10-15 | 2017-10-31 | 上海交通大学 | Compression video acquisition and reconfiguration system based on structural sparse dictionary learning |
| US10057383B2 (en) | 2015-01-21 | 2018-08-21 | Microsoft Technology Licensing, Llc | Sparsity estimation for data transmission |
| CN105721869B (en) * | 2016-01-26 | 2018-04-06 | 上海交通大学 | The collection of compression tensor and reconfiguration system based on structural sparse |
| CN110620927B (en) * | 2019-09-03 | 2022-05-27 | 上海交通大学 | Scalable compression video tensor signal acquisition and reconstruction system based on structured sparsity |
| CN110719473B (en) * | 2019-09-03 | 2021-11-23 | 上海交通大学 | Scalable compression video acquisition and reconstruction system based on structured sparsity |
| CN110944373B (en) * | 2019-09-27 | 2023-09-26 | 国家电网有限公司 | Wireless sensor network system, data transmission method, storage medium and terminal |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103237204A (en) * | 2013-03-21 | 2013-08-07 | 上海交通大学 | Video signal collection and reconfiguration system based on high-dimension compressed sensing |
-
2013
- 2013-09-16 CN CN201310422841.7A patent/CN103517079B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103237204A (en) * | 2013-03-21 | 2013-08-07 | 上海交通大学 | Video signal collection and reconfiguration system based on high-dimension compressed sensing |
Non-Patent Citations (3)
| Title |
|---|
| Distributed Video Coding Using Compressive Sampling;J. Prades-Nebot, Y. Ma, and T. Huang;《Proceedings of Picture Coding Symposium》;20090508;第1-4页 * |
| Robust Recovery of Signals From a Structured Union of Subspaces;Yonina C. Eldar and Moshe Mishali;《IEEE Transactions on Information Theory》;20091130;第II节部分B至第III节，Fig.1 * |
| 基于高维压缩感知的视频信号采集与重构技术研究;叶鑫炜;《中国优秀硕士学位论文全文数据库》;20130715;正文第9-17页第二章 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103517079A (en) | 2014-01-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103517079B (en) | Compression video acquisition based on data-driven subspace collection and reconfiguration system | |
| US10412324B2 (en) | Infrared image reconstruction method based on block-sparse compressive sensing and system thereof | |
| CN104301728A (en) | Compressed video capture and reconstruction system based on structured sparse dictionary learning | |
| CN103237204B (en) | Based on video signal collective and the reconfiguration system of higher-dimension compressed sensing | |
| CN105721869B (en) | The collection of compression tensor and reconfiguration system based on structural sparse | |
| Yang et al. | Distributed sensor perception via sparse representation | |
| CN104199627B (en) | Gradable video encoding system based on multiple dimensioned online dictionary learning | |
| CN110111256A (en) | Image Super-resolution Reconstruction method based on residual error distillation network | |
| CN105844635A (en) | Sparse representation depth image reconstruction algorithm based on structure dictionary | |
| CN108960333A (en) | Lossless compression method for high spectrum image based on deep learning | |
| CN109685737A (en) | A kind of image defogging method | |
| CN106204482A (en) | Based on the mixed noise minimizing technology that weighting is sparse | |
| CN105741333A (en) | Real-time compression and reconstruction method of Video-SAR (Synthetic Aperture Radar) image | |
| CN109982071A (en) | The bis- compression video detecting methods of HEVC based on time space complexity measurement and local prediction residual distribution | |
| CN107292821B (en) | A kind of super-resolution image reconstruction method and system | |
| CN103903271B (en) | Image forensics method for natural image and compressed and tampered image based on DWT | |
| CN106993240A (en) | Many video summarization methods based on sparse coding | |
| CN104243986B (en) | Compression video acquisition and reconfiguration system based on data-driven tensor subspace | |
| Xiong et al. | An unsupervised dictionary learning algorithm for neural recordings | |
| CN108288295A (en) | The method for fast reconstruction and system of infrared small target image based on structural information | |
| CN104881846A (en) | Structured image compressive sensing restoration method based on double-density dual-tree complex wavelet | |
| Wang et al. | Image reconstruction from patch compressive sensing measurements | |
| Xiao et al. | MFEN: Lightweight multi-scale feature extraction super-resolution network in embedded system | |
| Prasetyo et al. | Integrating Companding and Deep Learning on Bandwidth-Limited Image Transmission | |
| He et al. | MTRFN: Multiscale temporal receptive field network for compressed video action recognition at edge servers |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |