CN110136187A - A method of divided based on compressed sensing observing matrix and reduces relevance imaging computing cost - Google Patents
A method of divided based on compressed sensing observing matrix and reduces relevance imaging computing cost Download PDFInfo
- Publication number
- CN110136187A CN110136187A CN201910414311.5A CN201910414311A CN110136187A CN 110136187 A CN110136187 A CN 110136187A CN 201910414311 A CN201910414311 A CN 201910414311A CN 110136187 A CN110136187 A CN 110136187A
- Authority
- CN
- China
- Prior art keywords
- matrix
- compressed sensing
- light intensity
- distribution
- read
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/50—Depth or shape recovery
- G06T7/55—Depth or shape recovery from multiple images
- G06T7/557—Depth or shape recovery from multiple images from light fields, e.g. from plenoptic cameras
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10052—Images from lightfield camera
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Compression Of Band Width Or Redundancy In Fax (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of methods for dividing reduction relevance imaging computing cost based on compressed sensing observing matrix, comprising: the relevance imaging measurement data for obtaining experiment utilizes compressed sensing algorithm to carry out image reconstruction;A large amount of data are needed for reconstruct imaging, lead to the memory overflow problem occur when compressed sensing operation, compressed sensing is optimized and solved the problems, such as by trial.The results showed that this method can effectively reduce the memory overflow problem in compressed sensing restructing algorithm, and it can successfully reconstruct required image.
Description
Technical field
The invention belongs to compressed sensing based relevance imaging field, in particular to optimization of a kind of compressed sensing to memory
Method.
Background technique
Relevance imaging is the novel imaging mode of one kind of rising in recent years, the even High order correletion of the second order based on light field
Information carries out coincidence measurement by two or more detectors, realizes the association reconstruct of object to be measured.Relevance imaging technology because
Its special non-localized imaging property and its outstanding noise robustness have attracted the concern of a large number of researchers.Relevance imaging device
Structure is simple, has certain hyperresolution, so possessing boundless application prospect.
But want to obtain the reconstructed image of better quality in relevance imaging, it just has to acquire a large amount of experimental data,
And it takes a long time.The solution operator workload that will lead to the later period in this way is bigger than normal, and this be also relevance imaging development faced it is huge
Challenge.
And compressed sensing enters the sight of researcher, compressed sensing can alleviate this problem to a certain extent.
Compressed sensing compensates for the defect that original relevance imaging needs a large amount of data acquisition in signal acquisition process technology.Compression sense
Know and only needs the sampled data output more less than conventional Nyquist sampling processing data volume can very quickly and accurately
Restore original signal, improves the processing speed of information.Even so, but compressed sensing is applied in a large amount of relevance imaging data
The problem larger to memory requirements is equally faced with when processing, this is also a problem to be solved.
Summary of the invention
The purpose of the present invention is overcoming the deficiencies in the prior art, a kind of compressed sensing is provided and is reconstructed in relevance imaging
When the problem of needing a large amount of memories optimization method.
Technical solution provided by the invention are as follows:
A kind of compressed sensing applies the optimization method in relevance imaging, comprising:
The A matrix of compressed sensing is subjected to longitudinal choosing column according to demand, and carries out compressed sensing image reconstruction;
The A matrix of compressed sensing is subjected to laterally choosing row according to demand, and carries out compressed sensing image reconstruction.
It is associated imaging experiment, each frame reference arm light field needed for obtaining image reconstruction, and obtain corresponding
Bucket data;
The reference arm light field of first frame is read out, the distribution of light intensity matrix of a M*M is obtained;
Distribution of light intensity matrix is rotated clockwise 90 °, obtains postrotational distribution of light intensity matrix;
Preferably, obtained result arranges the longitudinal choosing for carrying out A matrix.It specifically includes:
Postrotational distribution of light intensity matrix is become into the one-dimension array comprising M*N element by reshape operation;
The N number of element of continuous P * (the continuous P row of postrotational light intensity matrix) in one-dimension array is intercepted according to demand;
The one-dimension array being truncated to is subjected to reshape operation, is reassembled as the distribution of light intensity matrix of P*N;
Obtained distribution of light intensity matrix is traversed and is written into new file;
Identical operation is carried out to each frame reference arm light field matrix and is written in corresponding new file according to serial number;
Preferably, the A matrix of compressed sensing carries out compressed sensing image reconstruction after longitudinally selecting column, how to carry out A matrix
Building, specifically include:
By treated, distribution of light intensity matrix file is read frame by frame;
The light field matrix that first frame light field reads to obtain P*N is subjected to reshape operation, is become comprising P*N member
The one-dimension array of element;
An empty A matrix is constructed, and A matrix is added as the first row in the one-dimension array after reshape;
Continue read the second frame distribution of light intensity matrix file, carry out reshape after as A matrix the second row.Into
Row as above all read and handle until by all distribution of light intensity matrix files by operation.
Preferably, the compressed sensing image reconstruction further include:
The bucket data that experiment measures are read, and are done normalized, and every a line of A matrix is done into normalizing
Change processing;
By the A matrix and bucket data progress compressed sensing image reconstruction after normalized, can be obtained expected
Reconstructed image.
Preferably, the compressed sensing A matrix laterally go by choosing, which is characterized in that goes back before carrying out laterally choosing row
Include:
Bucket data file is read, and is done Gauss curve fitting and obtains its mean value E and variances sigma;
The mean value E that Gauss curve fitting the obtains σ for subtracting n times is obtained into upper limit a, mean value E is obtained into lower limit b plus n times of σ.
Preferably, A matrix building and image reconstruction, which is characterized in that the place of distribution of light intensity matrix file
Reason includes:
First frame distribution of light intensity matrix file is read, and judges whether its corresponding bucket numerical value is less than a or is greater than
b;
If corresponding bucket data are unsatisfactory for requiring, give up and read next frame file;
If corresponding bucket data are met the requirements, the light field matrix of the M*N read is rotated clockwise 90 °, and will
Its reshape is at the one-dimension array comprising M*N element;
Using one-dimension array as the first row of A matrix, and using its corresponding bucket numerical value as one-dimension array bucket0
First element;
The second frame distribution of light intensity matrix file is read, is similarly judged, gives up if being unsatisfactory for requiring, meets the requirements
Then using its reshape as the second row of A matrix, and using its corresponding bucket numerical value as second element of bucket0;
As above operation is repeated, until all distribution of light intensity matrix files all read and handle.
The one-dimension array bucket0 of reconstruct is read, and is done normalized, and every a line of A matrix is done and is returned
One change processing;
By the A matrix and bucket0 data progress compressed sensing image reconstruction after normalized, can be obtained expected
Reconstructed image.
The present invention is include at least the following beneficial effects: after obtaining relevance imaging experimental data, being attempted using compression sense
Perception method carries out the reconstruct of image.When selecting column method using the longitudinal direction of A matrix to realize, can according to the size of selection at
Function reconstructs a part of ideal image, and can effectively reduce the data volume of computer reading, to reduce compression sense
Know the demand to calculator memory.When selecting row method using the transverse direction of A matrix to realize, can obtain contributing image information
The best part distribution of light intensity matrix, and success reconstructed image, to reduce demand of the compressed sensing to calculator memory.
Detailed description of the invention
Fig. 1 is that a kind of compressed sensing of the present invention applies the flow diagram in the optimization method of relevance imaging;
Fig. 2 is that compressed sensing A matrix of the present invention longitudinally selects column flow diagram;
What Fig. 3 illustrated is to carry out experimental data figure as restructuring procedure after longitudinal choosing arranges;
What Fig. 4 illustrated is the pre-treatment step before carrying out laterally choosing row to bucket data;
What Fig. 5 illustrated is to carry out laterally choosing row and image reconstruction procedure;
Fig. 6 difference n value tests reconstruct image, and it is 1.28,1.07,0.71,0.5,0.35,0.21,0.1 that n, which distinguishes value,;
The memory consumption that Fig. 7 .A matrix choosing row needs;
The image reconstruction result of Fig. 8 difference choosing column;
The memory consumption that Fig. 9 .A matrix choosing column need;
Figure 10: corresponding under different n values to select row image reconstruction signal-to-noise ratio;
Figure 11: difference choosing column operation image reconstruction SNR
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts all other
Embodiment shall fall within the protection scope of the present invention.
The advantages of to make technical solution of the present invention, is clearer, makees specifically to the present invention with reference to the accompanying drawings and examples
It is bright.
As shown in Figure 1, compressed sensing provided in an embodiment of the present invention applies the optimization method in relevance imaging, including following
Step:
It S1, take relevance imaging experimental data as input, the longitudinal of progress A matrix selects column, and new file is written;
Wherein, as shown in Fig. 2, the detailed process of step S1 are as follows:
S11, a frame relevance imaging laboratory reference arm light field is read first, obtain the distribution of light intensity matrix of M*M;
S12, distribution of light intensity matrix is rotated clockwise 90 °, obtains postrotational distribution of light intensity matrix;
S13, by postrotational distribution of light intensity matrix by reshape operation become one it is one-dimensional comprising M*N element
Array;
S14, the according to demand N number of element of continuous P * (the continuous P row of postrotational light intensity matrix) in interception one-dimension array;
S15, the one-dimension array being truncated to is subjected to reshape operation, is reassembled as the distribution of light intensity matrix of P*N;
S16, obtained distribution of light intensity matrix is traversed and is written into new file;
S17, identical operation is carried out to each frame reference arm light field matrix and corresponding new text is written according to serial number
In part;
S2, new file progress compressed sensing image reconstruction is read;
Wherein, as shown in figure 3, the detailed process of step S2 are as follows:
One S21, building empty A matrix;
S22, by treated, distribution of light intensity matrix file is read out;
S23, the light field matrix that first frame light field reads to obtain P*N is subjected to reshape operation, is become comprising P*N
The one-dimension array of a element;
S24, A matrix is added as current end row in the one-dimension array after reshape;
S25, judge whether file reading finish, and if it is unread finish, carry out the reading of next frame file and repeat with
Upper operation finishes if reading, and carries out in next step;
S26, bucket data are read, and normalized is done to it;
S27, each row of A matrix is normalized;
S28, by after normalized A matrix and bucket data carry out compressed sensing image reconstruction, can be obtained pre-
The reconstructed image of phase.
S3, the bucket data preprocessing operation before carrying out A matrix transverse direction choosing row are as follows;
Wherein, as shown in figure 5, step S3 method particularly includes:
S31, the bucket data that experiment measures are read;
S32, it bucket data is done into Gauss curve fitting obtains its mean value E and variances sigma;
S33, the mean value E that Gauss curve fitting the obtains σ for subtracting n times is obtained into upper limit a, mean value E is obtained down plus n times of σ
Limit b.
The building of S4, A matrix and image reconstruction, which is characterized in that the processing to distribution of light intensity matrix file includes.
Wherein, as shown in fig. 6, step S4 includes the following steps: to what distortion target was determined
S41, a frame distribution of light intensity matrix file is read, and judges whether its corresponding bucket numerical value is less than a or big
In b;
If S42, corresponding bucket data are unsatisfactory for requiring, give up and read next frame file;
If S43, corresponding bucket data are met the requirements, the light field matrix of the M*N read is rotated clockwise 90 °,
And by its reshape at the one-dimension array comprising M*N element;
S44, using one-dimension array as the first row of A matrix, and using its corresponding bucket numerical value as one-dimension array
First element of bucket0;
S45, the second frame distribution of light intensity matrix file is read, is similarly judged, given up if being unsatisfactory for requiring, met
It is required that then using its reshape as the current end row of A matrix, and using its corresponding bucket numerical value as the of bucket0
Two elements.
S46, as above operation is repeated, until all distribution of light intensity matrix files all read and handle.
S47, the one-dimension array bucket0 for reading reconstruct, and normalized is done, and every a line of A matrix is equal
Do normalized;
S48, by after normalized A matrix and bucket0 data carry out compressed sensing image reconstruction, can be obtained pre-
The reconstructed image of phase.
In order to verify the broad applicability of the embodiment of the present invention, the embodiment of the present invention has prepared two groups of experiments, has respectively corresponded
The longitudinal direction of row image reconstruction and A matrix is selected to select column image reconstruction in the transverse direction of A matrix.
The transverse direction of A matrix select row image reconstruction be according to relevance imaging experiment provide experimental data, the experiment be using
Capitalization GI is as target object.The lateral choosing row operation that experimental data is carried out to A matrix, then carries out image reconstruction, this
Testing reconstructed image should be a complete GI letter, this is first group of experiment.It is root that column image reconstruction is selected in the longitudinal direction of A matrix
Then longitudinal choosing column operation that the identical data that upper experiment provides accordingly carries out A matrix carries out image reconstruction, this experiment reconstruct image
As that should be part GI letter.
Therefore particular content when actually being tested of the embodiment of the present invention is as follows:
(1) A matrix laterally selects row image reconstruction to test
Relevance imaging experimental data totally 5000 frame distribution of light intensity file is inputted, the mean value E and variance of bucket are calculated
σ.Then the calculating for the bound chosen to bucket data is carried out.Upper limit a is enabled to be equal to E-n σ, lower limit b is equal to E+n σ.Become
The value of n is changed to carry out the selection of the row of A matrix.This experiment takes n to be equal to 1.28,1.07,0.71,0.5,0.35 respectively,
0.21,0.1 carries out seven groups of experiments, and corresponding to and being actually used in the frame number of reconstruct is respectively 617,998,1951,2905,3410,
4072,4503, experimental result is illustrated in fig. 6 shown below.Check during the experiment carry out operation required for calculator memory with
And the time, it is illustrated in fig. 7 shown below.Experimental result picture is subjected to SNR analysis with reference to figure, the results are shown in Figure 10.
(2) A matrix longitudinally selects column image reconstruction to test
Totally 5000 frame distribution of light intensity file, the choosing for carrying out A matrix arrange operation to input relevance imaging experimental data first, respectively
Choose preceding 1/2,1/3,1/4,1/5/, 1/6 five group of data of A matrix.After completing choosing column operation, data are saved into respectively new
In file.Then start to carry out compressed sensing image reconstruction experiment respectively, experimental result is illustrated in fig. 8 shown below.In experimentation
In check carry out operation required for calculator memory, be illustrated in fig. 9 shown below.Experimental result is spliced, then and with reference to figure
SNR analysis is carried out, as a result as shown in figure 11.
By above-mentioned experimental result it is found that the carry out compressed sensing A matrix choosing column operation that the embodiment of the present invention proposes can
Well when carrying out image reconstruction using a large amount of relevance imaging data, reduce the calculator memory of consumption, and successfully reconstruct
It is expected that image.It can be good at utilizing when carrying out the choosing column operation of compressed sensing A matrix and contribute image bigger data reconstruction
Image.Experimental data and operation time needed for operation capable of effectively being reduced.
Although the embodiments of the present invention have been disclosed as above, but its is not only in the description and the implementation listed
With it can be fully applied to various fields suitable for the present invention, for those skilled in the art, can be easily
Realize other modification, therefore without departing from the general concept defined in the claims and the equivalent scope, the present invention is simultaneously unlimited
In specific details and legend shown and described herein.
Claims (8)
1. a kind of divide the method for reducing relevance imaging computing cost based on compressed sensing observing matrix characterized by comprising
The A matrix of compressed sensing is split according to demand, and carries out compressed sensing image reconstruction.
2. compressed sensing A Factorization algorithm as described in claim 1 simultaneously carries out image reconstruction, comprising:
Compressed sensing A matrix is subjected to longitudinal choosing column according to demand, then carries out compressed sensing image reconstruction;
Compressed sensing A matrix is subjected to laterally choosing row according to demand, then carries out compressed sensing image reconstruction.
3. longitudinally choosing arranges compressed sensing A matrix as claimed in claim 2, which is characterized in that vertical carrying out compressed sensing A matrix
Before Xiang Xuanlie further include:
It is associated imaging experiment, each frame reference arm light field needed for obtaining image reconstruction, and obtain corresponding bucket number
According to;
The reference arm light field of first frame is read out, the distribution of light intensity matrix of a M*N is obtained;
Distribution of light intensity matrix is rotated clockwise 90 °, obtains postrotational distribution of light intensity matrix.
4. according to the method described in claim 2, its obtained result arranges the longitudinal choosing for carrying out A matrix.It specifically includes:
Postrotational distribution of light intensity matrix is become into the one-dimension array comprising M*N element by reshape operation;
The N number of element of continuous P * (the continuous P row of postrotational light intensity matrix) in one-dimension array is intercepted according to demand;
The one-dimension array being truncated to is subjected to reshape operation, is reassembled as the distribution of light intensity matrix of P*N;
Obtained distribution of light intensity matrix is traversed and is written into new file;
Identical operation is carried out to each frame reference arm light field matrix and is written in corresponding new file according to serial number.
5. the A matrix of compressed sensing as claimed in claim 2 carries out compressed sensing image reconstruction after longitudinally choosing arranges, feature exists
In how carrying out the building of A matrix, specifically include:
Construct an empty A matrix;
Distribution of light intensity matrix file after choosing column is read frame by frame;
The light field matrix that first frame light field reads to obtain P*N is subjected to reshape operation, is become to include P*N element
One-dimension array;
A matrix is added as the first row in one-dimension array after reshape;
Continue read the second frame distribution of light intensity matrix file, carry out reshape after as A matrix the second row.It carries out such as
Upper operation is all read and is handled until by all distribution of light intensity matrix files.
6. compressed sensing image reconstruction as claimed in claim 2, comprising:
The bucket data that experiment measures are read, and are done normalized, and every a line of A matrix is done at normalization
Reason;
By after normalized A matrix and bucket data carry out compressed sensing operation, expected reconstructed image can be obtained.
7. compressed sensing A matrix as claimed in claim 2 laterally go and carry out image reconstruction by choosing, which is characterized in that carrying out
Before transverse cuts further include:
Bucket data file is read, and is done Gauss curve fitting and obtains its mean value E and variances sigma;
The mean value E that Gauss curve fitting the obtains σ for subtracting n times is obtained into upper limit a, mean value E is obtained into lower limit b plus n times of σ.
8. A matrix as claimed in claim 2 laterally selects capable and image reconstruction, which is characterized in that distribution of light intensity matrix file
Processing include:
Construct an empty A matrix;
First frame distribution of light intensity matrix file is read, and judges whether its corresponding bucket numerical value is less than a or is greater than b;
If corresponding bucket data are unsatisfactory for requiring, give up and read next frame file;
If corresponding bucket data are met the requirements, the light field matrix of the M*N read is rotated clockwise 90 °, and by its
Reshape is at the one-dimension array comprising M*N element;
Using one-dimension array as the first row of A matrix, and using its corresponding bucket numerical value as the of one-dimension array bucket0
One element;
The second frame distribution of light intensity matrix file is read, is similarly judged, is given up if being unsatisfactory for requiring, meeting the requirements will
Second row of its reshape as A matrix, and using its corresponding bucket numerical value as second element of bucket0;
As above operation is repeated, until all distribution of light intensity matrix files all read and handle.
The one-dimension array bucket0 of reconstruct is read, and is done normalized, and every a line of A matrix is normalized
Processing;
By after normalized A matrix and bucket0 data carry out compressed sensing operation, expected reconstruct image can be obtained
Picture.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910414311.5A CN110136187B (en) | 2019-05-17 | 2019-05-17 | Method for reducing associated imaging calculation overhead based on compressed sensing observation matrix segmentation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910414311.5A CN110136187B (en) | 2019-05-17 | 2019-05-17 | Method for reducing associated imaging calculation overhead based on compressed sensing observation matrix segmentation |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110136187A true CN110136187A (en) | 2019-08-16 |
CN110136187B CN110136187B (en) | 2021-09-24 |
Family
ID=67575093
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910414311.5A Expired - Fee Related CN110136187B (en) | 2019-05-17 | 2019-05-17 | Method for reducing associated imaging calculation overhead based on compressed sensing observation matrix segmentation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110136187B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111193925A (en) * | 2019-12-25 | 2020-05-22 | 杭州中威电子股份有限公司 | Image compressed sensing coding and normalization method based on block vector inner product |
CN111551955A (en) * | 2020-06-22 | 2020-08-18 | 北京理工大学 | Bionic blocking ghost imaging method and system |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1484905A2 (en) * | 2003-06-06 | 2004-12-08 | GE Medical Systems Global Technology Company LLC | Image processing method and apparatus |
CN104159048A (en) * | 2014-07-24 | 2014-11-19 | 南京邮电大学 | Compressive sensing uniform weighting relevance imaging method for non-uniform light field |
CN106203374A (en) * | 2016-07-18 | 2016-12-07 | 清华大学深圳研究生院 | A kind of characteristic recognition method based on compressed sensing and system thereof |
CN107622515A (en) * | 2017-09-06 | 2018-01-23 | 郑州大学 | The physical re-organization method of squeezed light field |
CN109035282A (en) * | 2018-08-02 | 2018-12-18 | 吉林工程技术师范学院 | The thresholding method of Hadamard coded modulation relevance imaging |
US20190043269A1 (en) * | 2017-08-03 | 2019-02-07 | The University Of North Carolina At Chapel Hill | Methods, systems, and computer readable media for modeling garments using single view images |
-
2019
- 2019-05-17 CN CN201910414311.5A patent/CN110136187B/en not_active Expired - Fee Related
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1484905A2 (en) * | 2003-06-06 | 2004-12-08 | GE Medical Systems Global Technology Company LLC | Image processing method and apparatus |
CN104159048A (en) * | 2014-07-24 | 2014-11-19 | 南京邮电大学 | Compressive sensing uniform weighting relevance imaging method for non-uniform light field |
CN106203374A (en) * | 2016-07-18 | 2016-12-07 | 清华大学深圳研究生院 | A kind of characteristic recognition method based on compressed sensing and system thereof |
US20190043269A1 (en) * | 2017-08-03 | 2019-02-07 | The University Of North Carolina At Chapel Hill | Methods, systems, and computer readable media for modeling garments using single view images |
CN107622515A (en) * | 2017-09-06 | 2018-01-23 | 郑州大学 | The physical re-organization method of squeezed light field |
CN109035282A (en) * | 2018-08-02 | 2018-12-18 | 吉林工程技术师范学院 | The thresholding method of Hadamard coded modulation relevance imaging |
Non-Patent Citations (3)
Title |
---|
LUO BIN.ET AL: ""Orthonormalization method in ghost imaging"", 《OPTICS EXPRESS》 * |
任越美等: ""压缩感知及其图像处理应用研究进展与展望"", 《自动化学报》 * |
董小亮灯: ""压缩感知重构算法在"鬼"成像中的应用研究"", 《信号处理》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111193925A (en) * | 2019-12-25 | 2020-05-22 | 杭州中威电子股份有限公司 | Image compressed sensing coding and normalization method based on block vector inner product |
CN111193925B (en) * | 2019-12-25 | 2021-12-07 | 杭州中威电子股份有限公司 | Image compressed sensing coding and normalization method based on block vector inner product |
CN111551955A (en) * | 2020-06-22 | 2020-08-18 | 北京理工大学 | Bionic blocking ghost imaging method and system |
Also Published As
Publication number | Publication date |
---|---|
CN110136187B (en) | 2021-09-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Gürsoy et al. | TomoPy: a framework for the analysis of synchrotron tomographic data | |
Ludtke et al. | EMAN: semiautomated software for high-resolution single-particle reconstructions | |
CN101971045B (en) | Coil selection for parallel magnetic resonance imaging | |
CN107615089B (en) | Modeling and verification method for compressed sensing and MRI | |
CN111524064A (en) | Fluorescence microscopic image super-resolution reconstruction method based on deep learning | |
CN110136187A (en) | A method of divided based on compressed sensing observing matrix and reduces relevance imaging computing cost | |
CN1799498A (en) | Quick generalized partially parallel acquisition(GRAPA) image reconstruction algorithm for magnetic resonance imaging(MRI) | |
CN111832227B (en) | Shale gas saturation determination method, device and equipment based on deep learning | |
CN112884001B (en) | Automatic grading method and system for graphitization of carbon steel | |
CN109060592B (en) | Shale loss gas quantity obtaining method and device and terminal equipment | |
CN113112447A (en) | Tunnel surrounding rock grade intelligent determination method based on VGG convolutional neural network | |
CN1663526A (en) | Method for removing Gibbs ring pseudo-image generated in MRI reconstruction process | |
CN116071338A (en) | Method, device and equipment for detecting surface defects of steel plate based on YOLOX | |
CN115358285A (en) | Method, device and equipment for selecting key geological parameters of block to be surveyed | |
CN115410024A (en) | Power image defect detection method based on dynamic activation thermodynamic diagram | |
CN115373976A (en) | Insurance testing method and device, computer equipment and storage medium | |
Cheng et al. | Seismic data reconstruction based on multiscale attention deep learning | |
CN117315686A (en) | Oracle auxiliary decoding classification method and system based on classification model | |
CN101635046B (en) | Image processing method and device based on compute unified device architecture (CUDA) technology | |
CN113125374B (en) | Method, device and equipment for detecting REE content in carbonate type rare earth deposit sample | |
EP3755447A1 (en) | Determining conditions for purification of proteins | |
CN115081485A (en) | AI-based automatic analysis method for magnetic flux leakage internal detection data | |
Pan et al. | Astroconformer: The prospects of analysing stellar light curves with transformer-based deep learning models | |
Koeshidayatullah et al. | Is attention all geosciences need? Advancing quantitative petrography with attention-based deep learning | |
Goul et al. | Use of an expert subsystem in decision recognition channeling |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210924 |
|
CF01 | Termination of patent right due to non-payment of annual fee |