CN109448065B - Compressed sensing method based on gradient blocking self-adaptive measurement - Google Patents
Compressed sensing method based on gradient blocking self-adaptive measurement Download PDFInfo
- Publication number
- CN109448065B CN109448065B CN201811181710.3A CN201811181710A CN109448065B CN 109448065 B CN109448065 B CN 109448065B CN 201811181710 A CN201811181710 A CN 201811181710A CN 109448065 B CN109448065 B CN 109448065B
- Authority
- CN
- China
- Prior art keywords
- block
- image
- column
- blocks
- row
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T9/00—Image coding
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Abstract
The invention relates to a compressed sensing method based on gradient block self-adaptive measurement. The method comprises the following steps: taking non-overlapping basic blocks as dividing units of the image, calculating the smoothness of each basic block by utilizing gradients in the vertical direction and the horizontal direction of the image, dividing the image into image blocks with uneven sizes according to the smoothness, and calculating the measurement rate of each image block; respectively measuring each image block according to the image block size classification and selection measurement matrix to obtain a measurement value; the decoding end introduces vector included angles as similarity judgment standards, and adopts a non-local low-rank regularized compressed sensing reconstruction algorithm to reconstruct. The invention designs a non-uniform block self-adaptive measurement compressed sensing method, which ensures that a decoded image has robustness and better reconstruction effect.
Description
Technical Field
The invention relates to the field of computer image processing, in particular to a compressed sensing method based on gradient block self-adaptive measurement.
Background
Compressed sensing is a brand new signal processing framework that breaks through the nyquist sampling theorem. Information compression and signal reconstruction are two important components of compressed sensing. The information compression method is mainly divided into two types, namely, the whole image is compressed and the image is compressed after being segmented. The whole compressed sensing often needs to store a larger measurement matrix, occupies a larger memory, and has huge calculation amount.
For this reason, researchers have proposed a block compression method, which divides an image into small blocks of a specified size, and then compresses each small image using the same measurement matrix. For example, chinese patent document CN 106301384A, "a signal reconstruction method based on block compressed sensing": uniformly dividing an original signal into sub-blocks, performing sparse transformation on the sub-blocks, and filtering; observing the obtained sub-signals to obtain an observation vector; and restoring the sub-signals by using the observation vector and the measurement matrix, and then linearly combining the sub-signals to obtain a reconstructed signal. The partitioned compressed sensing has the advantages of smaller storage measurement matrix and simpler calculation of a reconstruction algorithm; but the algorithm is reconstructed separately for each sub-signal and then combined linearly, is not robust and is prone to reconstructed image blocking artifacts.
In the document Compressive sampling based on frequency saliency for remote sensing imaging, in order to improve the reconstruction effect of block compression sensing, the measurement rate is allocated according to the saliency information of the block, so that the reconstruction effect is effectively improved. But the OMP algorithm is adopted in the reconstruction process, priori knowledge of the image is not fully utilized, and the set sparsity value has a large influence on the reconstruction effect.
The existing block compressed sensing algorithm adopts uniform blocks, the calculation complexity is reduced by the blocks, but the original structure among the image blocks is destroyed, so that the block compressed sensing algorithm has block effect; so many scholars have proposed a method of smoothing the difference between restored blocks and removing the block effect, but this also results in breaking the image restoration structure while removing the block effect, reducing the image reconstruction effect.
The non-local low-rank regularized compressed sensing reconstruction algorithm proposed in the document Compressive Sensing via Nonlocal Low-Rank Regularization fully utilizes the non-local similarity of images to obtain a very good reconstruction effect, but the algorithm adopts overall uniform measurement at a compression end and does not fully utilize measurement information. In order to obtain more excellent reconstruction effect, the invention adopts the non-local low-rank method to reconstruct at the decoding end.
Disclosure of Invention
In order to solve the defects in the prior art, solve the problem of block compressed sensing reconstruction and improve the peak signal to noise ratio of the reconstructed image, the invention aims to: a compressed sensing method based on gradient block adaptive measurement is provided.
The technical scheme adopted for solving the technical problems is as follows: in order to reduce the contradiction between the deblocking effect and the reserved restoration structure of the blocking method, an uneven blocking strategy is provided by combining the image geometry, the area with a similar structure is divided into one block as much as possible, the whole blocking quantity of the image and the generation of the blocking effect are reduced, and the structural loss during the deblocking effect is further reduced. When in measurement, the self-adaptive measurement is carried out according to the smoothness of the uneven blocks, and under the condition that the overall measurement rate is unchanged, the measurement rate of the smooth area is reduced and the measurement rate of the complex area is increased in a self-adaptive manner according to the smoothness, so that the utilization rate of effective information is improved.
The method comprises the following steps:
(1) Inputting an image to be processed, wherein a digitizing matrix of the image to be processed is X epsilon R m×n M and n are the number of rows and columns of the image matrix respectively;
(2) Blocking the image, dividing the image into d uneven block images based on image gradient, and expanding all image blocks into column vectors x j Where j= {1,2,..d }, is combined into a column vector x= [ x ] 1 ；x 2 ；...；x d ]；
(3) The measurement rate M of each block is allocated according to the sampling rate M required by the image and the smoothness of each block j ,j＝{1,2,...,d}；
(4) Dividing blocks with the same element number into one class, generating a Gaussian random orthogonal matrix with the same size, and combining the measurement rate M of each block j Selecting the front round (M) j ×length(x j ) Row phi as measurement matrix phi j The method comprises the steps of carrying out a first treatment on the surface of the The data are measured to obtain the measurement y=Φx=diag (Φ) 1 ,φ 2 ,...,φ d )[x 1 ；x 2 ；...；x d ]；
(5) Reconstructing an original signal by using the measured value, the block measurement rate, the line segmentation position and the measurement matrix, solving by using a non-local low-rank regularized compressed sensing reconstruction algorithm, and outputting a reconstructed image;
x j representing column vectors which are formed by expanding the image block matrix after the block division according to columns, wherein j is the order of the image blocks, and d blocks are formed;
φ j represents x j Selecting the number of rows according to the sampling rate, wherein the number of columns is determined by the size of the image block;
y j generating a measurement of the column vector for each image block;
the block algorithm in the step (2) is as follows:
step 1: calculating an image X ε R m×n Vertical gradient matrix D v And a horizontal gradient matrix D h ：
Wherein X (a, b) represents the pixel values in row a and column b in the image matrix X, D v (a,b)、D h (a, b) represent the vertical and horizontal gradients in row a and column b, respectively, in the image matrix X.
Step 2: the image X is divided into basic blocks of r X r which are not overlapped with each other, r is a multiple of 2, and if r X r is less than r X r, 0 is used for filling.
Step 3: respectively to the vertical gradient matrix D v And a horizontal gradient matrix D h Pooling operation is carried out by taking the basic block as a unit to respectively obtain a vertical smooth matrixAnd horizontal smoothing matrix->Calculating the smoothness in the vertical and horizontal directions representing each basic block based on the gradient in the vertical direction and the gradient in the horizontal direction：
Wherein S is v (p,q)、S h (p, q) respectively representing the vertical direction smoothness and the horizontal direction smoothness of the p-th row and q-th column base block in units of base blocks; epsilon is a constant, the denominator is avoided to be zero, and when epsilon=1 is taken, S v (p,q),S h (p,q)∈(1,255]The larger the value, the better the smoothness.
Step 4: will S v The elements of each row of (1) are divided into two classes, and a threshold T is set v ，S v (i,j)＜T v Is a non-smooth block S v (i,j)≥T v For smooth block, record N v (i) For the number of i-th line smooth blocks, calculating a line segmentation parameter:
wherein L is v (i) The proportion of the i-th line smooth block is shown, and the number of flat sliders is used as a reference value for line segmentation because the image structure has regionality.
Step 5: according to a vertical smoothness matrix S v Dividing the line blocks from top to bottom for the basic block image, R v Storing the line index of the last line basic block in the line blocks, length (R v ) The final number of line blocks.
5.1: initial i=1, j=1, r v (j)＝1，σ v Being constant, the line block segmentation threshold is indicated.
5.2: i=i+1, and judging |l v (i)-L v (i-1)|＜σ v If so, the i-th basic block line is classified as including the i-1-th basic block line, if the smoothness in the vertical direction of the i-th basic block line and the i-1-th basic block line is within the threshold valueIn the same row block of the base block row, R v (j) =i; otherwise, if the vertical smoothness of the two basic block rows is not within the threshold, dividing the i basic block row into the next row block, where j=j+1, r v (j)＝i。
5.3: judgingIf true, it indicates that the line block division is completed, step 6 is executed, and if not, 5.2 is executed.
Step 6: for each divided row of blocks, according to a horizontal smoothness matrix S h Column division is performed from left to right, and the division method is similar to the row division method.
6.1: initial j=0, r v (0)＝0；
6.2: j=j+1, extract S h The j-th row block s h ，s h ＝S h (R v (j-1)+1:R v (j) (:), for s h The elements of each column of (2) are divided into two classes, and a threshold T is set h ，S h (p,q)＜T h Is a non-smooth block S h (p,q)≥T h For smooth block, record N h (k) Representation s h Calculating column segmentation parameters according to the number of the k-th column smoothing blocks:
wherein L is h (k) Representation s h The proportion of the k-th smooth block, R v Storing the line index of the last line basic block of the divided line blocks, R v (j)-R v (j-1) represents s h The number of basic blocks in each column is equal to s h Number of elements in each column.
6.3: for s h Column division, R h (j, k) storing a last column index of a kth column block dividing the jth row block. length (R) h (j:%) is s h The number of column blocks is different from one row block to another, and the number of divided column blocks is different from one row block to another.
6.4：Initial i=1, k=1, r h (j,k)＝1，σ h The column block threshold is partitioned as a constant.
6.5: i=i+1, and judging |l h (k)-L h (k-1)|＜σ h If so, the k basic block column is classified into the same column block containing the k-1 basic block column when the smoothness in the horizontal direction of the k basic block column and the k-1 basic block column in the j-th row block is within the threshold value range h (j, k) =i; otherwise, if the horizontal smoothness of the two basic block columns is not within the threshold, dividing the k basic block column into the next column of blocks, where k=k+1, r h (j,k)＝i。
6.6: judgingIf not, the column division representing the j-th row block is not completed, and 6.5 is executed; if true, judge j=length (R v ) If not, executing step 6.2, if yes, completing all the blocks and storing R h (j,k)。
The measurement rate in the step (3) adopts an adaptive allocation algorithm, and the algorithm is as follows:
step 1: the integral measurement rate M of the input image is calculated to calculate the image block x j Smoothness and of all base blocks in q = 1,.:
step 2: the measurement rate is allocated according to the smoothness of the subblock:
wherein M is 0 And > 0 is the block lowest measurement rate.
The step (4) adopts a non-local low-rank regularized compressed sensing reconstruction algorithm to reconstruct, and column vectors are recombined into images and then processed during reconstruction;
due to the adoption of the technical scheme, the invention has the beneficial effects that: by adopting the block compressed sensing image reconstruction algorithm, the block effect caused by block compressed sensing is effectively reduced; compared with the existing self-adaptive block measurement method, the method has higher peak signal-to-noise ratio and better image visual effect.
Drawings
FIG. 1 is a general flow chart of a compressed sensing method algorithm based on gradient blocking adaptive measurement according to the present invention;
FIG. 2 is a flow chart of an algorithm based on gradient non-uniform blocking of the present invention;
Detailed Description
The following describes a compressed sensing method based on gradient block adaptive measurement according to the present invention in detail with reference to the accompanying drawings and a typical embodiment, and the algorithm specifically includes the following parts:
the compression end compresses the image by using a gradient-based non-uniform block self-adaptive measurement method, and the steps are as follows:
inputting an image X epsilon R to be processed m×n Calculating gradients in the horizontal direction and the vertical direction of the image; carrying out pooling dimension reduction on the image gradient according to the size of the basic block to obtain the smoothness of the image basic block in the horizontal direction and the smoothness of the image basic block in the vertical direction; the method comprises the steps of dividing an image into row blocks from top to bottom according to the smoothness in the vertical direction, and then dividing each row block into column blocks from left to right according to the smoothness in the horizontal direction, so that uneven block division is completed. The method comprises the following specific steps:
step 1: calculating an image X ε R m×n Vertical gradient matrix D v And a horizontal gradient matrix D h ：
Wherein X (a, b) represents the pixel values in row a and column b in the image matrix X, D v (a,b)、D h (a, b) represent the vertical and horizontal gradients in row a and column b, respectively, in the image matrix X.
Step 2: the image X is divided into basic blocks of r X r which are not overlapped with each other, r is a multiple of 2, and if r X r is less than r X r, 0 is used for filling.
Step 3: respectively to the vertical gradient matrix D v And a horizontal gradient matrix D h Pooling operation is carried out by taking the basic block as a unit to respectively obtain a vertical smooth matrixAnd horizontal smoothing matrix->The smoothness in the vertical and horizontal directions representing each base block is calculated from the gradient in the vertical direction and the gradient in the horizontal direction, respectively:
wherein S is v (p,q)、S h (p, q) respectively representing the vertical direction smoothness and the horizontal direction smoothness of the p-th row and q-th column base block in units of base blocks; epsilon is a constant, the denominator is avoided to be zero, and when epsilon=1 is taken, S v (p,q),S h (p,q)∈(1,255]The larger the value, the better the smoothness.
Step 4: will S v The elements of each row of (1) are divided into two classes, and a threshold T is set v ，S v (i,j)＜T v Is a non-smooth block S v (i,j)≥T v For smooth block, record N v (i) For the number of i-th line smooth blocks, calculating a line segmentation parameter:
wherein L is v (i) The proportion of the i-th line smooth block is shown, and the number of flat sliders is used as a reference value for line segmentation because the image structure has regionality.
Step 5: according to a vertical smoothness matrix S v Dividing the line blocks from top to bottom for the basic block image, R v Storing the line index of the last line basic block in the line blocks, length (R v ) The final number of line blocks.
5.1: initial i=1, j=1, r v (j)＝1，σ v Being constant, the line block segmentation threshold is indicated.
5.2: i=i+1, and judging |l v (i)-L v (i-1)|＜σ v If so, the i-th basic block line is classified into the same line block containing the i-1-th basic block line when the smoothness in the vertical direction of the i-th basic block line and the i-1-th basic block line is within the threshold range, and R v (j) =i; otherwise, if the vertical smoothness of the two basic block rows is not within the threshold, dividing the i basic block row into the next row block, where j=j+1, r v (j)＝i。
5.3: judgingIf true, it indicates that the line block division is completed, step 6 is executed, and if not, 5.2 is executed.
Step 6: for each divided row of blocks, according to a horizontal smoothness matrix S h Column division is performed from left to right, and the division method is similar to the row division method.
6.1: initial j=0, r v (0)＝0；
6.2: j=j+1, extract S h The j-th row block s h ，s h ＝S h (R v (j-1)+1:R v (j) (:), for s h The elements of each column of (2) are divided into two classes, and a threshold T is set h ，S h (p,q)＜T h Is a non-smooth block，S h (p,q)≥T h For smooth block, record N h (k) Representation s h Calculating column segmentation parameters according to the number of the k-th column smoothing blocks:
wherein L is h (k) Representation s h The proportion of the k-th smooth block, R v Storing the line index of the last line basic block of the divided line blocks, R v (j)-R v (j-1) represents s h The number of basic blocks in each column is equal to s h Number of elements in each column.
6.3: for s h Column division, R h (j, k) storing a last column index of a kth column block dividing the jth row block. length (R) h (j:%) is s h The number of column blocks is different from one row block to another, and the number of divided column blocks is different from one row block to another.
6.4: initial i=1, k=1, r h (j,k)＝1，σ h The column block threshold is partitioned as a constant.
6.5: i=i+1, and judging |l h (k)-L h (k-1)|＜σ h If so, the k basic block column is classified into the same column block containing the k-1 basic block column when the smoothness in the horizontal direction of the k basic block column and the k-1 basic block column in the j-th row block is within the threshold value range h (j, k) =i; otherwise, if the horizontal smoothness of the two basic block columns is not within the threshold, dividing the k basic block column into the next column of blocks, where k=k+1, r h (j,k)＝i。
6.6: judgingIf not, the column division representing the j-th row block is not completed, and 6.5 is executed; if true, judge j=length (R v ) If not, executing step 6.2, if yes, completing all the blocks and storing R h (j,k)。
Then based on the compressed imageThe required sampling rate M and smoothness of each block allocate the measurement rate M of each block j J= {1,2,..d }, the specific steps are as follows:
step 1: the integral measurement rate M of the input image is calculated to calculate the image block x j Smoothness and of all base blocks in q = 1,.:
step 2: the measurement rate is allocated according to the smoothness of the subblock:
wherein M is 0 And > 0 is the block lowest measurement rate.
Finally dividing blocks with the same element number into one class, generating a Gaussian random orthogonal matrix with the same size, and combining the measurement rate M of each block j Selecting the front round (M) j ×length(x j ) Row phi as measurement matrix phi j The method comprises the steps of carrying out a first treatment on the surface of the The data are measured to obtain the measurement y=Φx=diag (Φ) 1 ,φ 2 ,...,φ d )[x 1 ；x 2 ；...；x d ]，；
The decompression end adopts a non-local low-rank regularized compressed sensing reconstruction algorithm, and comprises the following specific steps:
step 1: obtaining an initial estimated image x by adopting DCT soft threshold algorithm according to the measured value, the block measurement rate, the line segmentation position and the measurement matrix (0) ；
Step 2: initializing, k=1, x (k) ＝x (0) K represents the number of iterations. Step size 5 will be x 1 Divided into overlapping image reference blocks of size 6 x 6 and arranged in image x according to each reference block 1 Is searched for 44 most similar blocks. The invention adopts cosine included angle function of vector to judge the similarity judgment standard:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing image x (k) Column vector of the ith reference block, of->Representing image x (k) Column vector of image block of size 6 x 6, c represents image block +.>And reference block->The larger the value, the larger the similarity, with a maximum value of 1. Forming 45 similar block column vectors including the reference block into a similar block matrix X i Wherein X is i ＝R i x＝[R i0 x,R i1 x,...,R iq x]R represents an extraction block operation, q=45.
Step 3: establishing a model, and a similar block matrix X i The low rank model of (2) is:
wherein, the liquid crystal display device comprises a liquid crystal display device,is the variance of gaussian noise.
The regularization term for using the low-rank model as the reconstruction model is:
wherein eta and lambda are regularization parameters, and are set as constants during reconstruction; l (L) i Epsilon) represents rank substitution function, using kernel functionThe number is used as a rank-substitution function,
the method adopts an alternate solution method and is converted into:
step 4: solving for a matrix X of similar blocks i Low rank matrix L i The singular value contraction algorithm is adopted to obtain:
wherein, the liquid crystal display device comprises a liquid crystal display device,σ j representation->U ΣV T Is->Singular value decomposition matrix of (x) + ＝max{x,0}。
Step 5, solving the reconstructed imageJudging termination conditions, if the output reconstructed image is satisfied, otherwise, when mod (k, T) =0, searching 45 most similar blocks for each reference block again to form a similar block matrix X i Turning to step 3. Wherein (1)>Representing the operation of restoring the similarity block in the ith similarity block group to the original image position, a>
It is to be understood that the foregoing is merely one particular embodiment of the invention, and that the invention is not limited to the specific arrangements shown or described, but rather that the claims are to cover all modifications within the true spirit and scope of the invention.
Claims (2)
1. The compressed sensing method based on gradient block self-adaptive measurement is characterized by comprising the following steps of:
the method comprises the following steps:
(1) Inputting an image to be processed, wherein a digitizing matrix of the image to be processed is X epsilon R m×n M and n are the number of rows and columns of the image matrix respectively;
(2) Blocking the image, dividing the image into d non-uniform image blocks based on image gradient, and expanding all the image blocks into column vectors x j Where j= {1,2,..d }, is combined into a column vector x= [ x ] 1 ；x 2 ；...；x d ]；
(3) Allocating a measurement rate M of each block according to a sampling rate M required for compressing an image and smoothness of each block j ,j＝{1,2,...,d}；
(4) Dividing blocks with the same element number into one class, generating a Gaussian random orthogonal matrix with the same size, and combining the measurement rate M of each block j Selecting the front round (M) j ×length(x j ) Row phi as measurement matrix phi j The method comprises the steps of carrying out a first treatment on the surface of the The data are measured to obtain the measurement y=Φx=diag (Φ) 1 ,φ 2 ,...,φ d )[x 1 ；x 2 ；...；x d ]；
(5) Reconstructing an original signal by using the measured value, the block measurement rate, the line segmentation position and the measurement matrix, solving by using a non-local low-rank regularized compressed sensing reconstruction algorithm, and outputting a reconstructed image;
the block algorithm in the step (2) is as follows:
step 1: calculating an image X ε R m×n Vertical gradient matrix D v And a horizontal gradient matrix D h ：
Wherein X (a, b) represents the pixel values in row a and column b in the image matrix X, D v (a,b)、D h (a, b) represent the vertical and horizontal gradients in row a and column b, respectively, in the image matrix X;
step 2: dividing the image X into basic blocks of r multiplied by r which are not overlapped with each other, wherein r is a multiple of 2, and if the number of the basic blocks is less than r multiplied by r, filling with 0;
step 3: respectively to the vertical gradient matrix D v And a horizontal gradient matrix D h Pooling operation is carried out by taking the basic block as a unit to respectively obtain a vertical smooth matrixAnd horizontal smoothing matrix->The smoothness in the vertical and horizontal directions representing each base block is calculated from the gradient in the vertical direction and the gradient in the horizontal direction, respectively:
wherein S is v (p,q)、S h (p, q) respectively representing the vertical direction smoothness and the horizontal direction smoothness of the p-th row and q-th column base block in units of base blocks; epsilon is a constant, the denominator is avoided to be zero, and when epsilon=1 is taken, S v (p,q),S h (p,q)∈(1,255]The larger the value, the better the smoothness;
step 4: will S v The elements of each row of (1) are divided into two classes, and a threshold T is set v ，S v (i,j)＜T v Is a non-smooth block S v (i,j)≥T v For smooth block, record N v (i) For the number of i-th line smooth blocks, calculating a line segmentation parameter:
wherein L is v (i) Representing the proportion of the ith row smooth block, and adopting the number proportion of the flat sliding blocks as a reference value for row segmentation because the image structure has regionality;
step 5: according to a vertical smoothness matrix S v Dividing the line blocks from top to bottom for the basic block image, R v Storing the line index of the last line basic block in the line blocks, length (R v ) The number of the final row blocks;
5.1: initial i=1, j=1, r v (j)＝1，σ v Is constant and represents a line block segmentation threshold;
5.2: i=i+1, judge L v (i)-L v (i-1)＜σ v If so, the i-th basic block line is classified into the same line block containing the i-1-th basic block line when the smoothness in the vertical direction of the i-th basic block line and the i-1-th basic block line is within the threshold range, and R v (j) =i; otherwise, if the vertical smoothness of the two basic block rows is not within the threshold, dividing the i basic block row into the next row block, where j=j+1, r v (j)＝i；
5.3: judgingIf yes, the line block division is completed, step 6 is executed, and if not, 5.2 is executed;
step 6: for each divided row of blocks, according to a horizontal smoothness matrix S h Column segmentation is carried out from left to right, and the segmentation method is similar to the row segmentation method;
6.1: initial j=0, r v (0)＝0；
6.2: j=j+1, extract S h The j-th row block s h ，s h ＝S h (R v (j-1)+1:R v (j) (:), for s h The elements of each column of (2) are divided into two classes, and a threshold T is set h ，S h (p,q)＜T h Is a non-smooth block S h (p,q)≥T h For smooth block, record N h (k) Representation s h Calculating column segmentation parameters according to the number of the k-th column smoothing blocks:
wherein L is h (k) Representation s h The proportion of the k-th smooth block, R v Storing the line index of the last line basic block of the divided line blocks, R v (j)-R v (j-1) represents s h The number of basic blocks in each column is equal to s h The number of elements in each column;
6.3: for s h Column division, R h (j, k) storing a last column index of a kth column block dividing a jth row block; length (R) h (j:%) is s h The number of column blocks is different, and the number of the divided column blocks of each row of blocks is different because the smoothness of each row of blocks is different;
6.4: initial i=1, k=1, r h (j,k)＝1，σ h Dividing a column block threshold value for a constant;
6.5: i=i+1, judge L h (k)-L h (k-1)＜σ h If so, the k basic block column is classified into the same column block containing the k-1 basic block column when the smoothness in the horizontal direction of the k basic block column and the k-1 basic block column in the j-th row block is within the threshold value range h (j, k) =i; otherwise, if the horizontal smoothness of the two basic block columns is not within the threshold, dividing the k basic block column into the next column of blocks, where k=k+1, r h (j,k)＝i；
6.6: judgingIf not, the column division representing the j-th row block is not completed, and 6.5 is executed; if true, judge j=length (R v ) If not, executing step 6.2, if yes, completing all the blocks and storing R h (j,k)。
The measurement rate in the step (3) adopts an adaptive allocation algorithm, and the algorithm is as follows:
step 1: the integral measurement rate M of the input image is calculated to calculate the image block x j Smoothness and of all base blocks in q = 1,.:
step 2: the measurement rate is allocated according to the smoothness of the subblock:
wherein M is 0 And > 0 is the block lowest measurement rate.
2. The compressed sensing method based on gradient block adaptive measurement according to claim 1, wherein the step (5) uses a similar block judgment criterion in a non-local low-rank regularized compressed sensing reconstruction algorithm:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing image x (k) Column vector of the ith reference block, of->Representing image x (k) Column vector, c table, of image block of size 6 x 6Picture block->And reference block->The larger the value, the larger the similarity, with a maximum value of 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811181710.3A CN109448065B (en) | 2018-10-11 | 2018-10-11 | Compressed sensing method based on gradient blocking self-adaptive measurement |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811181710.3A CN109448065B (en) | 2018-10-11 | 2018-10-11 | Compressed sensing method based on gradient blocking self-adaptive measurement |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109448065A CN109448065A (en) | 2019-03-08 |
| CN109448065B true CN109448065B (en) | 2023-07-25 |
Family
ID=65545399
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811181710.3A Active CN109448065B (en) | 2018-10-11 | 2018-10-11 | Compressed sensing method based on gradient blocking self-adaptive measurement |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109448065B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110418137B (en) * | 2019-07-31 | 2021-05-25 | 东华大学 | Cross subset guided residual block set measurement rate regulation and control method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103440675A (en) * | 2013-07-30 | 2013-12-11 | 湖北工业大学 | Overall situation reconstitution optimization model construction method for image block compressed sensing |
| CN107154061A (en) * | 2017-05-09 | 2017-09-12 | 北京航宇天穹科技有限公司 | The regularization coding/decoding method that a kind of splits' positions are perceived |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7995649B2 (en) * | 2006-04-07 | 2011-08-09 | Microsoft Corporation | Quantization adjustment based on texture level |
-
2018
- 2018-10-11 CN CN201811181710.3A patent/CN109448065B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103440675A (en) * | 2013-07-30 | 2013-12-11 | 湖北工业大学 | Overall situation reconstitution optimization model construction method for image block compressed sensing |
| CN107154061A (en) * | 2017-05-09 | 2017-09-12 | 北京航宇天穹科技有限公司 | The regularization coding/decoding method that a kind of splits' positions are perceived |
Non-Patent Citations (2)
| Title |
|---|
| Block compressed sampling of image signals by saliency based adaptive partitioning;Siwang Zhou 等;《Multimedia Tools and Applications》;20171006;全文 * |
| 基于灰度共生矩阵的图像自适应分块压缩感知方法;杜秀丽等;《计算机科学》;20180815;第45卷(第8期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109448065A (en) | 2019-03-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107730451B (en) | Compressed sensing reconstruction method and system based on depth residual error network | |
| Liu et al. | Image restoration using total variation with overlapping group sparsity | |
| Li et al. | Best-buddy gans for highly detailed image super-resolution | |
| CN110490832B (en) | Magnetic resonance image reconstruction method based on regularized depth image prior method | |
| CN113808032B (en) | Multi-stage progressive image denoising algorithm | |
| Li et al. | FilterNet: Adaptive information filtering network for accurate and fast image super-resolution | |
| CN109003265B (en) | No-reference image quality objective evaluation method based on Bayesian compressed sensing | |
| CN106952317B (en) | Hyperspectral image reconstruction method based on structure sparsity | |
| CN106780338B (en) | Rapid super-resolution reconstruction method based on anisotropy | |
| CN104199627B (en) | Gradable video encoding system based on multiple dimensioned online dictionary learning | |
| Shi et al. | Multi-scale deep networks for image compressed sensing | |
| CN106709872A (en) | Quick image super-resolution reconstruction method | |
| CN113781308A (en) | Image super-resolution reconstruction method and device, storage medium and electronic equipment | |
| CN107154061B (en) | Regularized decoding method for block compressed sensing | |
| CN109448065B (en) | Compressed sensing method based on gradient blocking self-adaptive measurement | |
| CN109741258B (en) | Image super-resolution method based on reconstruction | |
| CN112837220B (en) | Method for improving resolution of infrared image and application thereof | |
| CN106559670A (en) | A kind of improved piecemeal video compress perception algorithm | |
| WO2017070841A1 (en) | Image processing method and apparatus | |
| CN113592728A (en) | Image restoration method, system, processing terminal and computer medium | |
| CN111640080B (en) | CS image denoising reconstruction method based on hyperspectral total variation | |
| CN110728728B (en) | Compressed sensing network image reconstruction method based on non-local regularization | |
| CN111669183B (en) | Compressed sensing sampling and reconstruction method, equipment and storage medium | |
| CN113362338B (en) | Rail segmentation method, device, computer equipment and rail segmentation processing system | |
| Xia et al. | A regularized tensor decomposition method with adaptive rank adjustment for compressed-sensed-domain background subtraction |
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 |