CN104992456A - Multi-scale matrix coding method - Google Patents
Multi-scale matrix coding method Download PDFInfo
- Publication number
- CN104992456A CN104992456A CN201510341516.7A CN201510341516A CN104992456A CN 104992456 A CN104992456 A CN 104992456A CN 201510341516 A CN201510341516 A CN 201510341516A CN 104992456 A CN104992456 A CN 104992456A
- Authority
- CN
- China
- Prior art keywords
- matrix
- pixel
- matrixes
- scale
- scale matrixes
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention discloses a multi-scale matrix coding method. The multi-scale matrix coding method comprises the following steps: inputting an initial coding matrix in the same size as an image, performing first-stage scale division on the initial matrix to obtain a plurality of first-stage scale matrixes, performing second-stage scale division on the first-stage matrixes to obtain a plurality of second-stage scale matrixes, performing local coding on the second-stage scale matrixes, and performing global coding on the first-stage scale matrixes on the basis of the local coding. Based on a block segmentation compressed sensing technology, the further division is performed on the basis of a segmented block, the coding matrixes obtaining more information of the segmented block can be designed according to dependencies between pixels, and, compared with an existing coding matrix, the coding matrix provided in the invention has the better reconstruction effect.
Description
Technical field
The present invention relates to a kind of matrix coder method, particularly a kind of multiple dimensioned matrix coder method.
Background technology
Tradition remote sensor must meet Shannon's sampling theorem in signal acquisition, and namely sample frequency must not lower than 2 of signal highest frequency times.Along with the development trend of remote sensing images high spatial resolution, high time resolution, high spectral coverage resolution, the remote sensor according to Shannon's sampling theorem design causes magnanimity sampled data, and it stores, transmission and the contradiction of data processing; In addition, this limits the contradiction of the low the effect also high investment causing remote sensor to manufacture and design and performance improved.For breaking through Shannon's sampling theorem to the restriction of traditional development of remote sensing, compressive sensing theory brings the change of signal sampling theory.Compressed sensing points out that carrying out non-self-adapting linear projection to original signal obtains observed reading, then utilizes openness priori from known less is more value, to reconstruct original signal by solving an optimization problem with high probability.
For overcoming the drawback gathering a large amount of pixel when digital camera (CCD or CMOS) is taken pictures, rice university of the U.S. have developed " single pixel camera ", this camera is a kind of brand-new camera structure, uses digital micromirror array to complete the optical computing of the linear projection of image on pseudorandom two value model.Utilize the sampling of single signal photon detector to obtain obtaining piece image than count many some recoveries of much less of image pixel, and have the adaptive ability of image wavelength, this adaptive ability is not available for traditional CCD and cmos imaging device.
But the DMD quantity that single pixel camera of rice university's development needs is too many, cause whole imaging system volume comparatively large, cost is higher.
Summary of the invention
The object of the present invention is to provide a kind of multiple dimensioned matrix coder method, solve the problem that existing imaging system volume is large, cost is high.
The technical solution realizing the object of the invention is: a kind of multiple dimensioned matrix coder method, comprising:
The first step, the initial code matrix of the sizes such as input and image;
Second step, carries out one-level partition of the scale to initial code matrix and obtains multiple one-level Scale Matrixes;
3rd step, carries out secondary partition of the scale to one-level Scale Matrixes and obtains multiple secondary Scale Matrixes, carry out local code to secondary Scale Matrixes;
4th step, overall situation coding is carried out to one-level Scale Matrixes in local code basis, obtains two-layer yardstick encoder matrix.
Compared with prior art, its remarkable advantage is in the present invention:
(1) the present invention adopts multiple dimensioned matrix coder method, significantly reduces the integrated cost of compressed sensing sampling section (camera lens and DMD).
(2) the present invention is based on splits' positions cognition technology, the basis of piecemeal further divides, according to the dependence between pixel, the encoder matrix obtaining this piecemeal more information can be designed, compare existing encoder matrix, quality reconstruction is better.
(3) the present invention can realize the high-resolution imaging at Large visual angle angle by the combination of different scale optical system.
(4) the present invention is by the division of two-stage yardstick, both ensure that the acquisition of effective information, and can realize Image Reconstruction fast again.
Accompanying drawing explanation
Fig. 1 is multiple dimensioned matrix coder method flow diagram of the present invention.
Fig. 2 is multiple dimensioned model partition schematic diagram of the present invention.
Fig. 3 is the two-stage partition of the scale schematic diagram of the encoder matrix of the embodiment of the present invention.
Fig. 4 is the schematic diagram that the secondary Scale Matrixes of the embodiment of the present invention converts row vector to.
Embodiment
Composition graphs 1, Fig. 2, a kind of multiple dimensioned matrix coder method, comprises the following steps:
The first step, the initial code matrix of the sizes such as input and image;
Second step, carries out to initial code matrix the one-level Scale Matrixes that one-level partition of the scale obtains multiple N*N pixel, wherein N=2
n, 2≤n≤6, the zero padding process of initial code matrix direction deficiency anyhow;
3rd step, carries out to one-level Scale Matrixes the secondary Scale Matrixes that secondary partition of the scale obtains multiple S*S pixel, wherein S=2
m, 1≤m≤5 and m<n, carry out local code to secondary Scale Matrixes; Be specially:
Step 3-1, carries out 01 distribution to the secondary Scale Matrixes of S*S pixel;
Step 3-2, converts the secondary Scale Matrixes of S*S pixel to 1*S by row
2row vector or S
2* the column vector of 1, namely from left to right count first row and be arranged in order from top to bottom, then secondary series arranges from top to bottom in turn, until obtain 1*S
2row vector or S
2* the column vector of 1; A wherein continuous N pixel value is selected to be row vector or the column vector of 1; Wherein M>=C.K.log (S
2/ K), the value of C is 1, K is degree of rarefication;
The continuous N a selected pixel value is that row vector or the column vector of 1 is reduced to the matrix of S*S pixel by step 3-3.
4th step, overall situation coding is carried out to one-level Scale Matrixes in local code basis, and obtain two-layer yardstick encoder matrix, detailed process is:
Step 4-1, judges whether there is the M*M matrix that pixel value is 0 in the one-level Scale Matrixes of the N*N pixel after local code, if exist, perform step 4-2, if do not exist, the one-level Scale Matrixes of this N*N pixel is two-layer yardstick encoder matrix;
Step 4-2, arranges the S*S XOR matrix that a pixel is 1, and secondary Scale Matrixes composition pixel value being to any one S*S pixel of the M*M matrix of 0 carries out XOR, obtains two-layer yardstick encoder matrix.
Below in conjunction with specific embodiment, the invention will be further described.
Embodiment
Composition graphs 3, the multiple dimensioned matrix coder method of the present embodiment comprises the following steps:
The first step, the initial code matrix of the sizes such as input and image;
In hardware sampling, each printing opacity position in encoder matrix be will with the alignment of image slices vegetarian refreshments so, in emulation experiment, be also the encoder matrix that will have an identical size;
Second step, carries out to initial code matrix the one-level Scale Matrixes that one-level partition of the scale obtains multiple 16*16;
3rd step, carries out to one-level Scale Matrixes the secondary Scale Matrixes that secondary partition of the scale obtains multiple 2*2, carries out local code to secondary Scale Matrixes; Be specially:
01 distribution is carried out to the secondary Scale Matrixes of 2*2;
Composition graphs 4, converts the row vector of 1*4 to by row by the secondary Scale Matrixes of 2*2, select wherein continuous 2 pixel values to be the row vector of 1;
The row vector being 1 by continuous 2 pixel values selected is reduced to the matrix of 2*2.
4th step, overall situation coding is carried out to one-level Scale Matrixes in local code basis, obtains two-layer yardstick encoder matrix:
Step 4-1, judges whether there is the matrix that 2*2 pixel value is 0 in the one-level Scale Matrixes of the 16*16 pixel after local code, if exist, perform step 4-2, if do not exist, the one-level Scale Matrixes of this 16*16 pixel is two-layer yardstick encoder matrix;
Step 4-2, a 2*2 is set and pixel be 1 XOR matrix, XOR is carried out to the secondary Scale Matrixes that composition 2*2 pixel value is any one 2*2 pixel of the matrix of 0, obtains two-layer yardstick encoder matrix.
The present invention is based on splits' positions cognition technology, the basis of piecemeal further divides, according to the dependence between pixel, can design the encoder matrix obtaining this piecemeal more information, existing encoder matrix of comparing, quality reconstruction is better.
Claims (7)
1. a multiple dimensioned matrix coder method, is characterized in that, comprise the following steps:
The first step, the initial code matrix of the sizes such as input and image;
Second step, carries out one-level partition of the scale to initial code matrix and obtains multiple one-level Scale Matrixes;
3rd step, carries out secondary partition of the scale to one-level Scale Matrixes and obtains multiple secondary Scale Matrixes, carry out local code to secondary Scale Matrixes;
4th step, overall situation coding is carried out to one-level Scale Matrixes in local code basis, obtains two-layer yardstick encoder matrix.
2. multiple dimensioned matrix coder method according to claim 1, is characterized in that, carry out one-level partition of the scale in second step to initial code matrix, obtains the one-level Scale Matrixes of multiple N*N pixel, wherein N=2
n, 2≤n≤6.
3. multiple dimensioned matrix coder method according to claim 2, is characterized in that, carries out the secondary Scale Matrixes that secondary partition of the scale obtains S*S pixel, wherein S=2 in the 3rd step to one-level Scale Matrixes
m, 1≤m≤5 and m<n.
4. multiple dimensioned matrix coder method according to claim 3, is characterized in that, carries out local code, be specially in the 3rd step to secondary Scale Matrixes:
Step 3-1, carries out 01 distribution to the secondary Scale Matrixes of S*S pixel;
Step 3-2, converts the secondary Scale Matrixes of S*S pixel to 1*S by row
2row vector, select a wherein continuous N pixel value to be the row vector of 1; Wherein M>=C.K.log (S
2/ K), C is constant, and K is degree of rarefication.
Step 3-3, the row vector being 1 by the continuous N a selected pixel value is reduced to the matrix of S*S pixel.
5. matrix coder method according to claim 3, is characterized in that, carries out local code, be specially in the 3rd step to secondary Scale Matrixes:
Step 3-1, carries out 01 distribution to the secondary Scale Matrixes of S*S pixel;
Step 3-2, converts the secondary Scale Matrixes of S*S pixel to S by row
2* the column vector of 1, selects a wherein continuous N pixel value to be the column vector of 1; Wherein M>=C.K.log (S
2/ K), C is constant, and K is degree of rarefication;
Step 3-3, the column vector being 1 by the continuous N a selected pixel value is reduced to the matrix of S*S pixel.
6. the multiple dimensioned matrix coder method according to claim 4 or 5, is characterized in that, C=1.
7. the multiple dimensioned matrix coder method according to claim 4 or 5, is characterized in that, the 4th step is specially:
Step 4-1, judges whether there is the M*M matrix that pixel value is 0 in the one-level Scale Matrixes of the N*N pixel after local code, if exist, perform step 4-2, if do not exist, the one-level Scale Matrixes of this N*N pixel is two-layer yardstick encoder matrix;
Step 4-2, arranges the S*S XOR matrix that a pixel is 1, and secondary Scale Matrixes composition pixel value being to any one S*S pixel of the M*M matrix of 0 carries out XOR, obtains two-layer yardstick encoder matrix.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510341516.7A CN104992456B (en) | 2015-06-18 | 2015-06-18 | Multiple dimensioned matrix coder method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510341516.7A CN104992456B (en) | 2015-06-18 | 2015-06-18 | Multiple dimensioned matrix coder method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104992456A true CN104992456A (en) | 2015-10-21 |
| CN104992456B CN104992456B (en) | 2018-05-01 |
Family
ID=54304263
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510341516.7A Active CN104992456B (en) | 2015-06-18 | 2015-06-18 | Multiple dimensioned matrix coder method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104992456B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108881732A (en) * | 2018-08-29 | 2018-11-23 | 哈尔滨工业大学（深圳） | Single pixel camera high-quality video imaging system based on double Scale Matrixes algorithms |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5768437A (en) * | 1992-02-28 | 1998-06-16 | Bri Tish Technology Group Ltd. | Fractal coding of data |
| CN102438102A (en) * | 2011-10-26 | 2012-05-02 | 西安电子科技大学 | Super-resolution imaging system based on compression coding aperture and imaging method thereof |
| US20130011051A1 (en) * | 2011-07-07 | 2013-01-10 | Lockheed Martin Corporation | Coded aperture imaging |
| CN104036519A (en) * | 2014-07-03 | 2014-09-10 | 中国计量学院 | Partitioning compressive sensing reconstruction method based on image block clustering and sparse dictionary learning |
| US20150003738A1 (en) * | 2013-07-01 | 2015-01-01 | Xerox Corporation | Adaptive quality image reconstruction via a compressed sensing framework |
| CN104574323A (en) * | 2015-02-03 | 2015-04-29 | 中国人民解放军国防科学技术大学 | Quick imaging method on basis of hierarchical models and guide priori for single-pixel cameras |
-
2015
- 2015-06-18 CN CN201510341516.7A patent/CN104992456B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5768437A (en) * | 1992-02-28 | 1998-06-16 | Bri Tish Technology Group Ltd. | Fractal coding of data |
| US20130011051A1 (en) * | 2011-07-07 | 2013-01-10 | Lockheed Martin Corporation | Coded aperture imaging |
| CN102438102A (en) * | 2011-10-26 | 2012-05-02 | 西安电子科技大学 | Super-resolution imaging system based on compression coding aperture and imaging method thereof |
| US20150003738A1 (en) * | 2013-07-01 | 2015-01-01 | Xerox Corporation | Adaptive quality image reconstruction via a compressed sensing framework |
| CN104036519A (en) * | 2014-07-03 | 2014-09-10 | 中国计量学院 | Partitioning compressive sensing reconstruction method based on image block clustering and sparse dictionary learning |
| CN104574323A (en) * | 2015-02-03 | 2015-04-29 | 中国人民解放军国防科学技术大学 | Quick imaging method on basis of hierarchical models and guide priori for single-pixel cameras |
Non-Patent Citations (4)
| Title |
|---|
| DAVID J. BRADY等: "Multiscale lens design", 《OPTICS EXPRESS》 * |
| JAMES E. FOWLER等: "MULTISCALE BLOCK COMPRESSED SENSING WITH SMOOTHED PROJECTED LANDWEBER RECONSTRUCTION", 《19TH EUROPEAN SIGNAL PROCESSING CONFERENCE》 * |
| JI-XIN LIU等: "Mass spectrum data processing based on compressed sensing recognition and sparse difference recovery", 《2012 9TH INTERNATIOANL CONFERENCE ON FUZZY SYSTEMS AND KNOWLEDGE DISCOVERY》 * |
| 刘佶鑫等: "多尺度分型压缩感知遥感成像方法", 《测绘学报》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108881732A (en) * | 2018-08-29 | 2018-11-23 | 哈尔滨工业大学（深圳） | Single pixel camera high-quality video imaging system based on double Scale Matrixes algorithms |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104992456B (en) | 2018-05-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Goldstein et al. | The stone transform: Multi-resolution image enhancement and compressive video | |
| CN104065891A (en) | Machine Vision 3d Line Scan Image Acquisition And Processing | |
| US10113870B2 (en) | Machine vision system for forming a digital representation of a low information content scene | |
| JP6482012B2 (en) | Readout integrated circuit with integrated compressed sensing | |
| US10284793B2 (en) | Machine vision system for forming a one dimensional digital representation of a low information content scene | |
| JP5822937B2 (en) | Method for processing an image in the invisible spectral region, corresponding camera and measuring device | |
| CN110557637A (en) | Apparatus for encoding image, apparatus for decoding image, and image sensor | |
| US10089719B2 (en) | Signal observation device and signal observation method | |
| US20240007608A1 (en) | Multi-Processor Support for Array Imagers | |
| Dadkhah et al. | Compressive sensing image sensors-hardware implementation | |
| Cameron et al. | Design considerations for the processing system of a CNN-based automated surveillance system | |
| Zhao et al. | Super resolution imaging based on a dynamic single pixel camera | |
| US20120092476A1 (en) | Methods and apparatus to form a wavelet representation of a pathology slide having glass and tissue regions | |
| CN107590841B (en) | System and method for compressing image data generated by a Computed Tomography (CT) imaging system | |
| CN104639834A (en) | Method and system for transmitting camera image data | |
| CN109327700B (en) | SEI transmission-based image coding and decoding time delay testing method | |
| US9615024B1 (en) | Multi-resolution compressive sensing image processing | |
| CN109102463B (en) | Super-resolution image reconstruction method and device | |
| CN106525236A (en) | Template calibration method of compression coding spectral imaging system | |
| CN104992456A (en) | Multi-scale matrix coding method | |
| CN103955956A (en) | Compressed-sensing-oriented image combination reconstruction method | |
| Wang et al. | A Lightweight Recurrent Aggregation Network for Satellite Video Super-Resolution | |
| Jerez et al. | Spectral image fusion for increasing the spatio-spectral resolution through side information | |
| York et al. | A comparison of polarization image processing across different platforms | |
| CN112036271B (en) | Pedestrian re-identification method, system, medium and terminal based on Kalman filtering |
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 |