CN103903261A - Spectrum image processing method based on partition compressed sensing - Google Patents
Spectrum image processing method based on partition compressed sensing Download PDFInfo
- Publication number
- CN103903261A CN103903261A CN201410111351.XA CN201410111351A CN103903261A CN 103903261 A CN103903261 A CN 103903261A CN 201410111351 A CN201410111351 A CN 201410111351A CN 103903261 A CN103903261 A CN 103903261A
- Authority
- CN
- China
- Prior art keywords
- spectral coverage
- phi
- spectrum picture
- spectral
- matrix
- 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
Images
Landscapes
- Investigating Or Analysing Materials By Optical Means (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention discloses a spectrum image processing method based on partition compressed sensing. The method mainly solves the problems of traditional spectrum image processing methods that sampling cost is high and coding complexity is high. The method comprises the steps of conducting classification and difference operation on a spectrum image according to inter-spectrum correlation coefficients, conducting sampling by means of the partition compressed sensing technology, improving a traditional measurement matrix, and recovering the original spectrum image by means of the compressed sensing reconstitution algorithm. The method solves the problems of traditional spectrum image processing methods that sampling cost is high and coding complexity is high, and can effectively improve the reconstitution quality of the spectrum image under the condition that the sampling rate is unchanged.
Description
Technical field
The invention belongs to picture signal process field, relate to a kind of spectrum picture disposal route, particularly a kind of based on piecemeal compressed sensing (BlockCompressiveSensing, BCS) spectrum picture disposal route, can realize the reconstruction quality that effectively improves spectrum picture under the condition of identical sampling rate.
Background technology
Remote sensing images can provide a large amount of science data and multidate informations to people according to the spectral response characteristics of different atural objects, have high using value.But be different from two dimensional image, and spectrum picture presents three-dimensional character, and data volume is very huge.
In traditional image processing process, compression occurs in data by after complete collecting, and has caused on the one hand the increase of system complexity, on the other hand, has wasted a large amount of sampling resources.Compressed sensing, as a kind of emerging technology that has great potential, has realized in sampled signal signal has been compressed, and can greatly reduce the sampling of sensor and assess the cost.For the spectroscopic data of magnanimity, traditional compressed sensing Technology Need takies higher computational resource, and fortunately, piecemeal compressed sensing technology is by reducing to a great extent energy consumption by the processing of view data piecemeal, reduce computation complexity and easily operation, be more suitable for on-line system.But piecemeal compressed sensing technology can not effectively be utilized the intrinsic characteristic of spectrum picture, thereby further reduce the sampling cost of sensor, the reconstruction quality of raising image.
Summary of the invention
For above problem, the present invention proposes a kind of spectrum picture disposal route based on piecemeal compressed sensing, Spectral correlation by spectrum picture is classified and difference operation to spectrum picture, utilize piecemeal compressed sensing technology to sample, and traditional perception matrix is weighted to processing, finally realize the reconstruction property that improves spectrum picture under the condition of identical sampling rate.
For achieving the above object, the spectrum picture disposal route based on piecemeal compressed sensing of the present invention, comprises the following steps:
(1) according to the Spectral correlation of spectrum picture, spectrum picture is carried out to Clustering:
1a) introduce related coefficient between spectrum, and determine the correlativity between different spectral coverage according to it:
Wherein, the horizontal total number-of-pixels that M is spectral coverage, longitudinal total number-of-pixels that N is spectral coverage, the horizontal ordinate that x is pixel, the ordinate that y is pixel, I (i, x, y) represents the pixel value that in i spectral coverage, coordinate (x, y) is located,
represent the average of pixel in i spectral coverage, r
i,jrepresent the related coefficient between i spectral coverage and j spectral coverage;
1b) select with the related coefficient of other spectral coverage and maximum spectral coverage as initial clustering point;
1c) by with step 1b) related coefficient of the cluster point the selected spectral coverage that is greater than threshold epsilon is divided into same group;
1d) repeating step 1b), 1c) until all spectral coverages are all classified, can be divided three classes through the spectral coverage of classification, with reference to spectral coverage, independent reconstruct spectral coverage and depend on the spectral coverage with reference to spectral coverage, represents with R, D and I respectively;
(2) successively the spectrum picture in same group is carried out to difference operation:
Wherein, Ψ is sparse base, and Φ is for measuring matrix, and Θ is the product of Ψ and Φ, x
kfor R class spectral coverage, x
ifor D class spectral coverage, x
ican rely on x
kbe reconstructed;
(3) utilize piecemeal compressed sensing technology to carry out piecemeal and improve traditional perception matrix the spectrum picture through difference operation:
3a) sub-block of B × B will be divided into through the spectrum picture of difference operation, place's zero padding that border is not enough;
3b) definition weight matrix w
b=[w
ij]
b × B, its element is w
ij=1+ (i+j-1) × q, and by this matrix-vector, obtain
q is a constant;
3c) according to w'
bimprove traditional perception matrix:
Wherein,
represent original perception matrix, m
brepresent perception value number, Φ
brepresent the perception matrix after improving;
(4) utilize the perception matrix after improving block data to be sampled and difference reconstruct.
The present invention has the following advantages compared with prior art:
1, the present invention carries out adaptive classification according to the Spectral correlation of spectrum picture by spectrum picture, without determining in advance clusters number, and can guarantee that the spectral coverage that correlativity is stronger is divided into same group.
2, the present invention, to carrying out difference operation through the spectrum picture of classification, and utilizes piecemeal compressed sensing technology to sample, and can effectively save system resource, and the low and more easily operation of encoder complexity, is applicable to on-line system.
3, the present invention, according to the importance of different frequency composition in image, is weighted processing to traditional perception matrix, can effectively improve the reconstruct effect of image, simple to operate, is easy to realize.
Accompanying drawing explanation
Fig. 1 is overview flow chart of the present invention.
Fig. 2 is the spectrum picture sorting technique process flow diagram based on Spectral correlation of the present invention.
Fig. 3 is in the situation that sampling rate is given, the PSNR performance simulation figure of distinct methods.
Embodiment
Describe the present invention referring to accompanying drawing:
With reference to Fig. 1, the present invention is based on the spectrum picture disposal route of piecemeal compressed sensing, comprise the following steps:
Step 1: spectrum picture is classified according to the Spectral correlation of spectrum picture.
With reference to Fig. 2, being implemented as follows of this step:
1a) introduce related coefficient between spectrum, and determine the correlativity between different spectral coverage according to it:
Wherein, the horizontal total number-of-pixels that M is spectral coverage, longitudinal total number-of-pixels that N is spectral coverage, the horizontal ordinate that x is pixel, the ordinate that y is pixel, I (i, x, y) represents the pixel value that in i spectral coverage, coordinate (x, y) is located,
represent the average of pixel in i spectral coverage, r
i,jrepresent the related coefficient between i spectral coverage and j spectral coverage;
1b) select with the related coefficient of other spectral coverage and maximum spectral coverage as initial clustering point;
1c) by with step 1b) related coefficient of the cluster point the selected spectral coverage that is greater than threshold epsilon is divided into same group;
1d) repeating step 1b), 1c) until all spectral coverages are all classified, can be divided three classes through the spectral coverage of classification, with reference to spectral coverage, independent reconstruct spectral coverage and depend on the spectral coverage with reference to spectral coverage, represents with R, D and I respectively.
Step 2: successively the spectrum picture in same group is carried out to difference operation.
Wherein, Ψ is sparse base, and Φ is for measuring matrix, and Θ is the product of Ψ and Φ, x
kfor R class spectral coverage, x
ifor D class spectral coverage, x
ican rely on x
kbe reconstructed.
Step 3: utilize piecemeal compressed sensing technology to carry out piecemeal and improve traditional perception matrix the spectrum picture through difference operation.
3a) sub-block of B × B will be divided into through the spectrum picture of difference operation, place's zero padding that border is not enough;
3b) definition weight matrix w
b=[w
ij]
b × B, its element is w
ij=1+ (i+j-1) × q, and by this matrix-vector, obtain
q is a constant;
3c) according to w'
bimprove traditional perception matrix:
Wherein,
represent original perception matrix, m
brepresent perception value number, Φ
brepresent the perception matrix after improving.
Step 4: utilize the perception matrix after improving block data to be sampled and difference reconstruct.
Effect of the present invention can further illustrate by following simulation result:
1, experiment condition:
The present embodiment is the Realization of Simulation under Matlab language environment, adopt respectively DCT matrix, Gauss's matrix as sparse base and perception matrix, dividing block size is 16, reconstructing method is OMP, the sampling rate of R class, D class and I class spectral coverage is respectively 0.4,0.2 and 0.3, and the sampling rate of independent reconstructing method is 0.3.Emulation adopt high spectrum test image from AVIRIS, be respectively the 6th scape and the 5th scape of LowAltitude, for convenience's sake, use respectively HS1 and HS2 mark they.Meanwhile, multispectral test pattern is MS1 and MS2, and they are 8 bit image and are made up of 7 spectral coverages, are respectively Rio, Paris.For for simplicity, test pattern all uses 64 × 64 subgraph.Rule of thumb, parameter q is taken as 3.
2, experiment content
A. utilize different perception matrixes to be reconstructed spectrum picture, its result as shown in Figure 3, wherein:
Fig. 3 (a) is the PSNR performance simulation figure of multispectral image Pairs;
Fig. 3 (b) is the PSNR performance simulation figure of spectral coverage 39~45 in high spectrum image LowAltitude the 6th scape;
Fig. 3 shows, for each spectral coverage, the PSNR performance of institute of the present invention extracting method is all obviously better than adopting the not BCS method of the perception matrix of weighting processing of tradition.Compare the method, the extracting method PSNR of institute of the present invention gain is the highest can reach 8.4dB, and minimum also have a 4.3dB.Visible, the quality of perception matrix is very large on the reconstruction property impact of compressed sensing, and institute of the present invention extracting method can effectively improve compressed sensing reconstruction property.
B. utilize distinct methods to be reconstructed spectral coverage in HS1 41~44, the size of test pattern is 256 × 256 here, and a point block size is 32 × 32, and its result is as shown in table 1:
Table 1PSNR performance (dB of unit)
Wherein, spectral coverage 42~44 is all larger with spectral coverage 41 correlativitys, and therefore they can rely on spectral coverage 41 and carry out difference reconstruct.Table 2 shows, for each spectral coverage, the inventive method can effectively improve the reconstruct effect of image compared with low sampling rate in the situation that, and wherein the independent reconstructing method performance based on weight is taken second place, the poorest without the performance of weight independence reconstructing method.
C. utilize distinct methods to divide into groups and difference operation to spectrum picture, perception matrix is the perception matrix based on weight that the present invention proposes here, and its PSNR performance is as shown in table 2:
Table 2PSNR performance (dB of unit)
Wherein, the mean P SNR that PSNR is each spectral coverage, and the average sample rate of high spectrum image and multispectral image is respectively 0.22 and 0.29.Table 1 shows, in the situation that sampling rate is identical, the PSNR performance of the inventive method is better, and the highest-gain of its relatively independent reconstructing method is greater than 4dB, the minimum 1dB that reaches, and its relative high spectrum image that gains is more remarkable.
The above, it is only preferred embodiment of the present invention, not the present invention is done to any pro forma restriction, although the present invention discloses as above with preferred embodiment, but not in order to limit the present invention, any those skilled in the art, do not departing within the scope of technical solution of the present invention, when can utilizing the method for above-mentioned announcement and technology contents to make a little change or being modified to the equivalent embodiment of equivalent variations, in every case be the content that does not depart from technical solution of the present invention, any simple modification of above embodiment being done according to technical spirit of the present invention, equivalent variations and modification, still belong in the scope of technical solution of the present invention.
Claims (4)
1. the spectrum picture disposal route based on piecemeal compressed sensing, is characterized in that, comprises the steps:
(1) according to the Spectral correlation of spectrum picture, spectrum picture is carried out to Clustering;
(2) successively the spectrum picture in same group is carried out to difference operation;
(3) utilize based on piecemeal compressed sensing technology to the spectrum picture through difference operation is carried out piecemeal and improves traditional perception matrix;
(4) utilize the perception matrix after improving block data to be sampled and difference reconstruct.
2. spectrum picture disposal route as claimed in claim 1, is characterized in that, described step (1) is as follows:
1a) introduce related coefficient between spectrum, and determine the correlativity between different spectral coverage according to it:
Wherein, the horizontal total number-of-pixels that M is spectral coverage, longitudinal total number-of-pixels that N is spectral coverage, the horizontal ordinate that x is pixel, the ordinate that y is pixel, I (i, x, y) represents the pixel value that in i spectral coverage, coordinate (x, y) is located,
represent the average of pixel in i spectral coverage, r
i,jrepresent the related coefficient between i spectral coverage and j spectral coverage;
1b) select with the related coefficient of other spectral coverage and maximum spectral coverage as initial clustering point;
1c) by with step 1b) related coefficient of the cluster point the selected spectral coverage that is greater than threshold epsilon is divided into same group;
1d) repeating step 1b), 1c) until all spectral coverages are all classified, can be divided three classes through the spectral coverage of classification, with reference to spectral coverage, independent reconstruct spectral coverage and depend on the spectral coverage with reference to spectral coverage, represents with R, D and I respectively.
3. spectrum picture disposal route as claimed in claim 1, is characterized in that, described step (2) is as follows:
Wherein, Ψ is sparse base, and Φ is for measuring matrix, and Θ is the product of Ψ and Φ, x
kfor R class spectral coverage, x
ifor D class spectral coverage, x
ican rely on x
kbe reconstructed.
4. spectrum picture disposal route as claimed in claim 1, is characterized in that, described step (3) is as follows:
3a) sub-block of B × B will be divided into through the spectrum picture of difference operation, place's zero padding that border is not enough;
3b) definition weight matrix w
b=[w
ij]
b × B, its element is w
ij=1+ (i+j-1) × q, and by this matrix-vector, obtain
q is a constant;
3c) according to w'
bimprove traditional perception matrix:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410111351.XA CN103903261B (en) | 2014-03-24 | 2014-03-24 | Spectrum image processing method based on partition compressed sensing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410111351.XA CN103903261B (en) | 2014-03-24 | 2014-03-24 | Spectrum image processing method based on partition compressed sensing |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103903261A true CN103903261A (en) | 2014-07-02 |
CN103903261B CN103903261B (en) | 2017-04-12 |
Family
ID=50994568
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410111351.XA Expired - Fee Related CN103903261B (en) | 2014-03-24 | 2014-03-24 | Spectrum image processing method based on partition compressed sensing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103903261B (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
CN104779960A (en) * | 2015-03-20 | 2015-07-15 | 南京邮电大学 | A signal reconstruction method based on block compressed sensing |
CN104778665A (en) * | 2015-04-14 | 2015-07-15 | 清华大学 | Compressed ghost imaging reconstruction method based on natural image block prior driving and system |
CN105354867A (en) * | 2015-11-27 | 2016-02-24 | 中国矿业大学(北京) | Hyperspectral image compression algorithm research of adaptive redundant dictionary compressed sensing |
CN105631906A (en) * | 2015-12-23 | 2016-06-01 | 百度在线网络技术(北京)有限公司 | Texture compression method and device of three-dimensional map image |
CN106780636A (en) * | 2016-11-14 | 2017-05-31 | 深圳大学 | The sparse reconstructing method and device of a kind of image |
CN107133992A (en) * | 2017-04-17 | 2017-09-05 | 东北大学 | Based on spectrum correlation adaptive grouping high spectrum image distributed associating compression method |
CN108133500A (en) * | 2017-12-22 | 2018-06-08 | 杭州电子科技大学 | A kind of splits' positions reconstructing method towards plant EO-1 hyperion |
CN108460777A (en) * | 2017-12-22 | 2018-08-28 | 杭州电子科技大学 | A kind of extraction splits' positions reconstructing method towards plant EO-1 hyperion |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102034250A (en) * | 2010-11-26 | 2011-04-27 | 西安电子科技大学 | Edge structure information based block compression perception reconstruction method |
US20130016284A1 (en) * | 2011-07-12 | 2013-01-17 | Xerox Corporation | Hyperspectral image reconstruction via a compressed sensing framework |
-
2014
- 2014-03-24 CN CN201410111351.XA patent/CN103903261B/en not_active Expired - Fee Related
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102034250A (en) * | 2010-11-26 | 2011-04-27 | 西安电子科技大学 | Edge structure information based block compression perception reconstruction method |
US20130016284A1 (en) * | 2011-07-12 | 2013-01-17 | Xerox Corporation | Hyperspectral image reconstruction via a compressed sensing framework |
Non-Patent Citations (2)
Title |
---|
JINGYUAN LV等: "Hyperspectral compressive sensing", 《PROC.OF SPIE》 * |
李蕴华: "一种改进的图像分块压缩感知模型", 《计算机工程与应用》 * |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
CN104036519B (en) * | 2014-07-03 | 2017-05-10 | 中国计量学院 | Partitioning compressive sensing reconstruction method based on image block clustering and sparse dictionary learning |
CN104779960B (en) * | 2015-03-20 | 2018-04-03 | 南京邮电大学 | A kind of signal reconfiguring method perceived based on splits' positions |
CN104779960A (en) * | 2015-03-20 | 2015-07-15 | 南京邮电大学 | A signal reconstruction method based on block compressed sensing |
CN104778665A (en) * | 2015-04-14 | 2015-07-15 | 清华大学 | Compressed ghost imaging reconstruction method based on natural image block prior driving and system |
CN104778665B (en) * | 2015-04-14 | 2018-12-11 | 清华大学 | Compression ghost image reconstruction method and system based on the driving of natural image block priori |
CN105354867A (en) * | 2015-11-27 | 2016-02-24 | 中国矿业大学(北京) | Hyperspectral image compression algorithm research of adaptive redundant dictionary compressed sensing |
CN105631906A (en) * | 2015-12-23 | 2016-06-01 | 百度在线网络技术(北京)有限公司 | Texture compression method and device of three-dimensional map image |
CN106780636A (en) * | 2016-11-14 | 2017-05-31 | 深圳大学 | The sparse reconstructing method and device of a kind of image |
CN106780636B (en) * | 2016-11-14 | 2020-06-12 | 深圳大学 | Sparse reconstruction method and device for image |
CN107133992A (en) * | 2017-04-17 | 2017-09-05 | 东北大学 | Based on spectrum correlation adaptive grouping high spectrum image distributed associating compression method |
CN107133992B (en) * | 2017-04-17 | 2019-07-12 | 东北大学 | Based on spectrum correlation adaptive grouping high spectrum image distributed associating compression method |
CN108133500A (en) * | 2017-12-22 | 2018-06-08 | 杭州电子科技大学 | A kind of splits' positions reconstructing method towards plant EO-1 hyperion |
CN108460777A (en) * | 2017-12-22 | 2018-08-28 | 杭州电子科技大学 | A kind of extraction splits' positions reconstructing method towards plant EO-1 hyperion |
CN108133500B (en) * | 2017-12-22 | 2019-07-16 | 杭州电子科技大学 | A kind of splits' positions reconstructing method towards plant EO-1 hyperion |
CN108460777B (en) * | 2017-12-22 | 2020-06-05 | 杭州电子科技大学 | Extraction, blocking, compression and reconstruction method for hyperspectrum of plant |
Also Published As
Publication number | Publication date |
---|---|
CN103903261B (en) | 2017-04-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103903261A (en) | Spectrum image processing method based on partition compressed sensing | |
CN105163121B (en) | Big compression ratio satellite remote sensing images compression method based on depth autoencoder network | |
CN111274865A (en) | Remote sensing image cloud detection method and device based on full convolution neural network | |
CN106408530A (en) | Sparse and low-rank matrix approximation-based hyperspectral image restoration method | |
CN105160678A (en) | Convolutional-neural-network-based reference-free three-dimensional image quality evaluation method | |
CN105208374A (en) | Non-reference image quality objective evaluation method based on deep learning | |
CN104199627B (en) | Gradable video encoding system based on multiple dimensioned online dictionary learning | |
CN110827198B (en) | Multi-camera panoramic image construction method based on compressed sensing and super-resolution reconstruction | |
CN103037212B (en) | The adaptive block compressed sensing method for encoding images of view-based access control model perception | |
CN105761290A (en) | Adaptive multi-scale partitioning compression sensing sampling method | |
CN105049851A (en) | Channel no-reference image quality evaluation method based on color perception | |
CN110533591B (en) | Super-resolution image reconstruction method based on codec structure | |
CN106022355A (en) | 3DCNN (three-dimensional convolutional neural network)-based high-spectral image space spectrum combined classification method | |
CN104680180A (en) | Polarimetric SAR image classification method on basis of K-Means and sparse own coding | |
CN107133992A (en) | Based on spectrum correlation adaptive grouping high spectrum image distributed associating compression method | |
CN105550712A (en) | Optimized convolution automatic encoding network-based auroral image sorting method | |
CN102148986A (en) | Method for encoding progressive image based on adaptive block compressed sensing | |
CN104036519A (en) | Partitioning compressive sensing reconstruction method based on image block clustering and sparse dictionary learning | |
CN103220016B (en) | Generation system and method of pulse ultra wideband system orthogonal sparse dictionary | |
CN116306780B (en) | Dynamic graph link generation method | |
CN117253043A (en) | Remote sensing image semantic segmentation method based on spatial information enhanced transducer | |
CN110472732B (en) | Image feature extraction system based on optimized feature extraction device | |
CN112215241A (en) | Image feature extraction device based on small sample learning | |
CN116309221A (en) | Method for constructing multispectral image fusion model | |
CN104036509B (en) | Method for unmixing hyperspectral mixed pixel based on compressed sensing |
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 | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170412 |
|
CF01 | Termination of patent right due to non-payment of annual fee |