CN107392975A - The multiple dimensioned splits' positions cognitive method of image adaptive, electronic equipment - Google Patents
The multiple dimensioned splits' positions cognitive method of image adaptive, electronic equipment Download PDFInfo
- Publication number
- CN107392975A CN107392975A CN201710621894.XA CN201710621894A CN107392975A CN 107392975 A CN107392975 A CN 107392975A CN 201710621894 A CN201710621894 A CN 201710621894A CN 107392975 A CN107392975 A CN 107392975A
- Authority
- CN
- China
- Prior art keywords
- image
- block
- flat
- texture
- adaptive
- 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
- G06T11/00—2D [Two Dimensional] image generation
- G06T11/001—Texturing; Colouring; Generation of texture or colour
Abstract
The present invention, which discloses a kind of multiple dimensioned splits' positions cognitive method of image adaptive and electronic equipment, method, to be included:Small echo area image is obtained after image to be compressed is carried out into wavelet decomposition;High frequency wavelet area image is obtained after wavelet inverse transformation to estimate texture image;After texture image progress self-adaptive processing will be estimated, texture image data after being handled;Low frequency wavelet area image is obtained after wavelet inverse transformation to estimate Flat image;After Flat image progress self-adaptive processing will be estimated, Flat image data after being handled;Flat image data after texture image data after processing and processing are reconstructed, obtain texture image after reconstructing back edge image and reconstructing, texture image is overlapped to obtain reconstruction image after reconstructing back edge image and reconstruct.The present invention is rebuild high frequency as texture block self-adjusted block sample rate, low frequency as flat block, so as to fully remove blocking effect.
Description
Technical field
The present invention relates to image procossing correlative technology field, particularly a kind of multiple dimensioned splits' positions of image adaptive perceive
Method, electronic equipment and storage medium.
Background technology
Compressed sensing concept has turned into the focus in information research field since proposition, due in compressive sensing theory, believing
The sampling and compression of breath are carried out simultaneously, reduce memory space and amount of calculation and sample rate breaches sample rate in Shannon's theorems
More than or equal to the limitation of twice of bandwidth, sample rate is reduced, that is, passes through less sampling can quick reconfiguration signal.
However, there is following defect in existing compressed sensing algorithm:
First kind algorithm, compressed sensing algorithm is combined with wavelet transformation, each image subblock is individually rebuild, obtained
Obtained preferable reconstruction effect.However, due to using identical sample rate to each image block, unreasonable point of resource is caused
Match somebody with somebody, and blocking effect can be produced in process of reconstruction.
Second class algorithm, different sample rates is set to each layer of wavelet decomposition, i.e., using different observing matrixes to each
Layer wavelet coefficient is observed, and using smooth projection Landweber (SPL) algorithm in process of reconstruction, i.e., carries out wiener in spatial domain
Filtering, threshold process is carried out in transform domain, reconstruction quality is improved, eliminates blocking effect.However, due to just for wavelet decomposition
Low frequency coefficient afterwards is rebuild, so the image reconstruction effect more complicated to details is undesirable, still suffers from obvious blocking effect.
The content of the invention
Based on this, it is necessary to for the technology of obvious blocking effect after conventional images compressed sensing algorithm image reconstruction be present
Problem, there is provided a kind of image adaptive multiple dimensioned splits' positions cognitive method, electronic equipment and storage medium.
The present invention provides a kind of multiple dimensioned splits' positions cognitive method of image adaptive, including:
Small echo area image is obtained after image to be compressed is carried out into wavelet decomposition;
High frequency wavelet area image will be obtained after the low frequency coefficient zero setting of the small echo area image, by the high frequency wavelet domain figure
As obtaining estimating texture image after wavelet inverse transformation;
By it is described estimate texture image carry out self-adaptive processing after, texture image data after being handled;
Low frequency wavelet area image will be obtained after the high frequency coefficient zero setting of the small echo area image, by the low frequency wavelet domain figure
As obtaining estimating Flat image after wavelet inverse transformation;
By it is described estimate Flat image carry out self-adaptive processing after, Flat image data after being handled;
Texture image data after the processing and Flat image data after the processing are reconstructed, obtain reconstructing back
Texture image after edge image and reconstruct, texture image after the reconstruct back edge image and the reconstruct is overlapped to obtain weight
Build image.
Further:
It is described by it is described estimate texture image carry out self-adaptive processing after, texture image data after handle, specifically wrap
Include:By it is described estimate texture image carry out the adaptively sampled inspection process of entropy after, texture image data after being handled;
It is described by it is described estimate Flat image carry out self-adaptive processing after, Flat image data after handle, specifically wrap
Include:The Flat image of estimating is subjected to the adjacent block edge adaptive weighted filter processing of two dimension, Flat image number after being handled
According to.
Further, it is described by it is described estimate texture image carry out the adaptively sampled inspection process of entropy after, handled
Texture image data afterwards, specifically include:
The texture image of estimating is divided into the image block that h size is B × B and non-overlapping copies;
According to formulaTry to achieve the adaptive sample rate r of every image blockj, wherein,
S is default target sampling rate, and h is piecemeal number, SminFor Least sampling rate threshold value, HjFor the gray level entropy of each image block;
According to formula Mj=rj×B2Calculate j-th piece of adaptive observation number Mj;
From the random matrix Φ of B × B orthogonal dimensions conversionB×BIn randomly select MjIndividual row vector forms observing matrix Φj;
Utilize observing matrix ΦjObtain observation set yjAs texture image data after processing.
It is further, described that the Flat image of estimating is subjected to the adjacent block edge adaptive weighted filter processing of two dimension,
Flat image data after being handled, are specifically included:
In described B pixel of symmetric extension of four direction up and down for estimating Flat image, estimated after being expanded
Flat image;
B for estimating Flat image after expansioni,jImage block, by Bi,jFour adjacent block B of image blocki-1,j、Bi+1,j、
Bi,j-1、Bi,j+1In, with Bi,jThe group of image block its adjacent a line, a row and four summits composition one (B+2) × (B+2)
Close block Bi,j', by combination block Bi,j' four summit zero setting;
To Bi,jB × B the pixel at ' center carries out smothing filtering, wherein:
To the pixel value b to be updated of the sub-block of the i-th row jth rowi,jSmoothly, obtain the sub-block renewal of the i-th row jth row
Pixel value b afterwardsi,j' be:
bi,j'=pc·bi,j+pi,j l·bi,0+pi,j r·bi,B+2+pi,j u·b0,j+pi,j d·bB+2,j, 1≤i, j≤B;
pc=0.5,p1=1/ (j)2,p2=1/ (B+1-j)2,p3=1/ (i)2,p4=1/ (B+1-i)2;
S=(p1+p2+p3+p4)·2;
pi,j l=p1/s,pi,j r=p2/s,pi,j u=p3/s,pi,j d=p4/ s, wherein:
pc、pi,j l、pi,j r、pi,j u、pi,j dFor five groups of smooth parameters, bi,0、bi,B+2、b0,j、bB+2,jFor pixel value to be updated
bi,jFour corresponding edge pixel values.
Further, small echo area image is obtained after the progress wavelet decomposition by image to be compressed, specifically included:It will treat
Compression image obtains small echo area image after carrying out three layers of wavelet decomposition.
The present invention provides a kind of electronic equipment, including:
At least one processor;And
The memory being connected with least one processor communication;Wherein,
The memory storage has can be by the instruction of one computing device, and the instruction is by least one place
Manage device to perform, so that at least one processor can:
Small echo area image is obtained after image to be compressed is carried out into wavelet decomposition;
High frequency wavelet area image will be obtained after the low frequency coefficient zero setting of the small echo area image, by the high frequency wavelet domain figure
As obtaining estimating texture image after wavelet inverse transformation;
By it is described estimate texture image carry out self-adaptive processing after, texture image data after being handled;
Low frequency wavelet area image will be obtained after the high frequency coefficient zero setting of the small echo area image, by the low frequency wavelet domain figure
As obtaining estimating Flat image after wavelet inverse transformation;
By it is described estimate Flat image carry out self-adaptive processing after, Flat image data after being handled;
Texture image data after the processing and Flat image data after the processing are reconstructed, obtain reconstructing back
Texture image after edge image and reconstruct, texture image after the reconstruct back edge image and the reconstruct is overlapped to obtain weight
Build image.
Further:
It is described by it is described estimate texture image carry out self-adaptive processing after, texture image data after handle, specifically wrap
Include:By it is described estimate texture image carry out the adaptively sampled inspection process of entropy after, texture image data after being handled;
It is described by it is described estimate Flat image carry out self-adaptive processing after, Flat image data after handle, specifically wrap
Include:The Flat image of estimating is subjected to the adjacent block edge adaptive weighted filter processing of two dimension, Flat image number after being handled
According to.
Further, it is described by it is described estimate texture image carry out the adaptively sampled inspection process of entropy after, handled
Texture image data afterwards, specifically include:
The texture image of estimating is divided into the image block that h size is B × B and non-overlapping copies;
According to formulaTry to achieve the adaptive sample rate r of every image blockj, wherein,
S is default target sampling rate, and h is piecemeal number, SminFor Least sampling rate threshold value, HjFor the gray level entropy of each image block;
According to formula Mj=rj×B2Calculate j-th piece of adaptive observation number Mj;
From the random matrix Φ of B × B orthogonal dimensions conversionB×BIn randomly select MjIndividual row vector forms observing matrix Φj;
Utilize observing matrix ΦjObtain observation set yjAs texture image data after processing.
It is further, described that the Flat image of estimating is subjected to the adjacent block edge adaptive weighted filter processing of two dimension,
Flat image data after being handled, are specifically included:
In described B pixel of symmetric extension of four direction up and down for estimating Flat image, estimated after being expanded
Flat image;
B for estimating Flat image after expansioni,jImage block, by Bi,jFour adjacent block B of image blocki-1,j、Bi+1,j、
Bi,j-1、Bi,j+1In, with Bi,jThe group of image block its adjacent a line, a row and four summits composition one (B+2) × (B+2)
Close block Bi,j', by combination block Bi,j' four summit zero setting;
To Bi,jB × B the pixel at ' center carries out smothing filtering, wherein:
To the pixel value b to be updated of the sub-block of the i-th row jth rowi,jSmoothly, obtain the sub-block renewal of the i-th row jth row
Pixel value b afterwardsi,j' be:
bi,j'=pc·bi,j+pi,j l·bi,0+pi,j r·bi,B+2+pi,j u·b0,j+pi,j d·bB+2,j, 1≤i, j≤B;
pc=0.5, p1=1/ (j)2,p2=1/ (B+1-j)2,p3=1/ (i)2,p4=1/ (B+1-i)2;
S=(p1+p2+p3+p4)·2;
pi,j l=p1/s,pi,j r=p2/s,pi,j u=p3/s,pi,j d=p4/ s, wherein:
pc、pi,j l、pi,j r、pi,j u、pi,j dFor five groups of smooth parameters, bi,0、bi,B+2、b0,j、bB+2,jFor pixel value to be updated
bi,jFour corresponding edge pixel values.
It is further again, small echo area image is obtained after the progress wavelet decomposition by image to be compressed, is specifically included:Will
Image to be compressed obtains small echo area image after carrying out three layers of wavelet decomposition.
The present invention provides a kind of storage medium, and the storage medium stores computer instruction, when computer performs the meter
When calculation machine instructs, for performing all steps of the multiple dimensioned splits' positions cognitive method of image adaptive as previously described.
The present invention is rebuild high frequency as texture block self-adjusted block sample rate, low frequency as flat block, so as to fully go
Except blocking effect.
Brief description of the drawings
Fig. 1 is a kind of workflow diagram of the multiple dimensioned splits' positions cognitive method of image adaptive of the present invention;
Fig. 2 is a kind of workflow for the multiple dimensioned splits' positions cognitive method of image adaptive that one embodiment of the invention provides
Cheng Tu;
Fig. 3 is the schematic diagram of the adjacent block edge adaptive weighted filter processing of one embodiment of the invention two dimension;
Fig. 4 is the workflow diagram of preferred embodiment;
Fig. 5 is the hardware structure diagram of a kind of electronic equipment of the present invention.
Embodiment
The present invention will be further described in detail with specific embodiment below in conjunction with the accompanying drawings.
Embodiment one
It is as shown in Figure 1 a kind of workflow diagram of the multiple dimensioned splits' positions cognitive method of image adaptive of the present invention, bag
Include:
Step S101, small echo area image is obtained after image to be compressed is carried out into wavelet decomposition;
Step S102, high frequency wavelet area image will be obtained after the low frequency coefficient zero setting of the small echo area image, by the height
Frequency small echo area image obtains estimating texture image after wavelet inverse transformation;
Step S103, by it is described estimate texture image carry out self-adaptive processing after, texture image data after being handled;
Step S104, low frequency wavelet area image will be obtained after the high frequency coefficient zero setting of the small echo area image, will be described low
Frequency small echo area image obtains estimating Flat image after wavelet inverse transformation;
Step S105, by it is described estimate Flat image carry out self-adaptive processing after, Flat image data after being handled;
Step S106, texture image data after the processing and Flat image data after the processing are reconstructed, obtained
Texture image after to reconstruct back edge image and reconstruct, texture image after the reconstruct back edge image and the reconstruct is carried out
Superposition obtains reconstruction image.
Specifically, the low frequency component of image is mainly the comprehensive measurement to whole sub-picture intensity, and HFS is then
To image border and the measurement of profile, therefore image carries out obtaining small echo area image after wavelet decomposition in step S101, to it
HFS performs step S102 and step S103 processing, and low frequency part is directly considered as into flat block, performs step S104
With step S105 processing, reconstructed finally by step S106 using such as SPL algorithms, the image after reconstruct is overlapped
Super-resolution reconstruction image is can obtain, because low frequency part is directly considered as flat block, therefore can fully rebuild block edge.
The present invention is rebuild high frequency as texture block self-adjusted block sample rate, low frequency as flat block, so as to fully go
Except blocking effect.
Embodiment two:
A kind of multiple dimensioned splits' positions cognitive method of image adaptive of one embodiment of the invention offer is provided
Workflow diagram, including:
Step S201, small echo area image is obtained after image to be compressed is carried out into three layers of wavelet decomposition.
Step S202, high frequency wavelet area image will be obtained after the low frequency coefficient zero setting of the small echo area image, by the height
Frequency small echo area image obtains estimating texture image after wavelet inverse transformation.
Step S203, by it is described estimate texture image carry out the adaptively sampled inspection process of entropy after, texture after being handled
View data, specifically include:The texture image of estimating is divided into the image block that h size is B × B and non-overlapping copies;
According to formulaTry to achieve the adaptive sample rate r of every image blockj, wherein,
S is default target sampling rate, and h is piecemeal number, SminFor Least sampling rate threshold value, HjFor the gray level entropy of each image block;
According to formula Mj=rj×B2Calculate j-th piece of adaptive observation number Mj;
From the random matrix Φ of B × B orthogonal dimensions conversionB×BIn randomly select MjIndividual row vector forms observing matrix Φj;
Utilize observing matrix ΦjObtain observation set yjAs texture image data after processing.
Specifically, according to formulaTry to achieve the adaptively sampled of every image block
Rate rj, and according to formula Mj=rj×B2Calculate each piece of adaptive observation number Mj, from the random of B × B orthogonal dimensions conversion
Matrix ΦB×BIn randomly select MjIndividual row vector forms observing matrix Φj, utilize formula(wherein s ∈ H, V,
D }, 1≤l≤L) obtain observation set yj, the l layers of wherein subscript l expression wavelet decompositions, to estimate texture image, it includes x
Horizontal component H, vertical component V and diagonal components D.
Step S204, low frequency wavelet area image will be obtained after the high frequency coefficient zero setting of the small echo area image, will be described low
Frequency small echo area image obtains estimating Flat image after wavelet inverse transformation;
Step S205, the Flat image of estimating is subjected to the adjacent block edge adaptive weighted filter processing of two dimension, obtained everywhere
Flat image data after reason, are specifically included:
In described B pixel of symmetric extension of four direction up and down for estimating Flat image, estimated after being expanded
Flat image;
B for estimating Flat image after expansioni,jImage block, by Bi,jFour adjacent block B of image blocki-1,j、Bi+1,j、
Bi,j-1、Bi,j+1In, with Bi,jThe group of image block its adjacent a line, a row and four summits composition one (B+2) × (B+2)
Close block Bi,j', by combination block Bi,j' four summit zero setting;
To Bi,jB × B the pixel at ' center carries out smothing filtering, wherein:
To the pixel value b to be updated of the sub-block of the i-th row jth rowi,jSmoothly, obtain the sub-block renewal of the i-th row jth row
Pixel value b afterwardsi,j' be:
bi,j'=pc·bi,j+pi,j l·bi,0+pi,j r·bi,B+2+pi,j u·b0,j+pi,j d·bB+2,j, 1≤i, j≤B;
pc=0.5, p1=1/ (j)2,p2=1/ (B+1-j)2,p3=1/ (i)2,p4=1/ (B+1-i)2;
S=(p1+p2+p3+p4)·2;
pi,j l=p1/s,pi,j r=p2/s,pi,j u=p3/s,pi,j d=p4/ s, wherein:
pc、pi,j l、pi,j r、pi,j u、pi,j dFor five groups of smooth parameters, bi,0、bi,B+2、b0,j、bB+2,jFor pixel value to be updated
bi,jFour corresponding edge pixel values.
Specifically, as shown in figure 3, to Bi,jB × B the pixel at ' center carries out smothing filtering, it is assumed that to the i-th row jth
The sub-block of row is carried out smoothly, because the edge pixel gray scale between adjacent block produces blocking effect caused by identical saltus step, therefore
The pixel, which is filtered, need to utilize the local edge of four adjacent blocks around this pixel.The edge pixel values of adjacent block are entered
Weighted sum of the row based on distance, obtains the pixel value b after renewali,j′。
Step S206, texture image data after the processing and Flat image data after the processing are reconstructed, obtained
Texture image after to reconstruct back edge image and reconstruct, texture image after the reconstruct back edge image and the reconstruct is carried out
Superposition obtains reconstruction image.
Embodiment three
The workflow diagram of preferred embodiment is illustrated in figure 4, including:
Step S401, edge expansion is carried out to original image;
Step S402,3 grades of wavelet decompositions are then carried out, the data after decomposition, perform step S403 and step S406 respectively;
Step S403, HFS zero setting is post-processed to obtain pre- reconstructed image F;
Step S404, the adaptively sampled inspection process of entropy will be performed after pre- reconstructed image F piecemeals;
Step S405, it will perform after the data after the adaptively sampled inspection process of entropy carry out SPL reconstruct and obtain reconstructing back
Edge image, perform step S409;
Step S406, low frequency part zero setting is post-processed to obtain pre- reconstructed image T;
Step S407, the adjacent block edge adaptive weighted filter of two dimension will be performed after pre- reconstructed image T piecemeals and is handled;
Step S408, the data after the adjacent block edge adaptive weighted filter processing of two dimension are subjected to SPL reconstruct, perform step
S409;
Step S409, edge image and texture image after reconstruct, which are overlapped, can obtain super-resolution reconstruction image.
Example IV
The hardware structure diagram of a kind of electronic equipment of the present invention is illustrated in figure 5, including:
At least one processor 501;And
The memory 502 communicated to connect with least one processor 501;Wherein,
The memory 502 is stored with can be by the instruction of one computing device, and the instruction is by described at least one
Individual computing device, so that at least one processor can:
Small echo area image is obtained after image to be compressed is carried out into wavelet decomposition;
High frequency wavelet area image will be obtained after the low frequency coefficient zero setting of the small echo area image, by the high frequency wavelet domain figure
As obtaining estimating texture image after wavelet inverse transformation;
By it is described estimate texture image carry out self-adaptive processing after, texture image data after being handled;
Low frequency wavelet area image will be obtained after the high frequency coefficient zero setting of the small echo area image, by the low frequency wavelet domain figure
As obtaining estimating Flat image after wavelet inverse transformation;
By it is described estimate Flat image carry out self-adaptive processing after, Flat image data after being handled;
Texture image data after the processing and Flat image data after the processing are reconstructed, obtain reconstructing back
Texture image after edge image and reconstruct, texture image after the reconstruct back edge image and the reconstruct is overlapped to obtain weight
Build image.
In Fig. 5 by taking a processor 502 as an example.
Server can also include:Input unit 503 and output device 505.
Processor 501, memory 502, input unit 503 and display device 505 can pass through bus or other modes
Connect, in figure exemplified by being connected by bus.
Memory 502 is used as a kind of non-volatile computer readable storage medium storing program for executing, available for storage non-volatile software journey
Sequence, non-volatile computer executable program and module, such as the multiple dimensioned piecemeal pressure of image adaptive in the embodiment of the present application
Programmed instruction/module corresponding to contracting cognitive method, for example, the method flow shown in Fig. 1, Fig. 2, Fig. 4.Processor 501 passes through fortune
Row is stored in non-volatile software program, instruction and module in memory 502, so as to perform various function application and number
According to processing, that is, realize the multiple dimensioned splits' positions cognitive method of image adaptive in above-described embodiment.
Memory 502 can include storing program area and storage data field, wherein, storing program area can store operation system
Application program required for system, at least one function;Storage data field can be stored according to the multiple dimensioned splits' positions of image adaptive
Cognitive method uses created data etc.., can be with addition, memory 502 can include high-speed random access memory
Including nonvolatile memory, for example, at least a disk memory, flush memory device or other non-volatile solid state memories
Part.In certain embodiments, memory 502 is optional including relative to the remotely located memory of processor 501, these are remotely deposited
Reservoir can pass through network connection to the device for performing the multiple dimensioned splits' positions cognitive method of image adaptive.The reality of above-mentioned network
Example includes but is not limited to internet, intranet, LAN, mobile radio communication and combinations thereof.
The user that input unit 503 can receive input clicks on, and produces and the multiple dimensioned splits' positions sense of image adaptive
The signal input that the user of perception method is set and function control is relevant.Display device 505 may include the display devices such as display screen.
It is stored in one or more of modules in the memory 502, when by one or more of processing
When device 501 is run, the multiple dimensioned splits' positions cognitive method of image adaptive in above-mentioned any means embodiment is performed.
Embodiment five
The a kind of electronic equipment that one embodiment of the invention provides, including:
At least one processor;And
The memory being connected with least one processor communication;Wherein,
The memory storage has can be by the instruction of one computing device, and the instruction is by least one place
Manage device to perform, so that at least one processor can:
Small echo area image is obtained after image to be compressed is carried out into three layers of wavelet decomposition.
High frequency wavelet area image will be obtained after the low frequency coefficient zero setting of the small echo area image, by the high frequency wavelet domain figure
As obtaining estimating texture image after wavelet inverse transformation.
By it is described estimate texture image carry out the adaptively sampled inspection process of entropy after, texture image data after being handled,
Specifically include:The texture image of estimating is divided into the image block that h size is B × B and non-overlapping copies;
According to formulaTry to achieve the adaptive sample rate r of every image blockj, wherein,
S is default target sampling rate, and h is piecemeal number, SminFor Least sampling rate threshold value,jFor the gray level entropy of each image block;
According to formula Mj=rj×B2Calculate each piece of adaptive observation number Mj;
From the random matrix Φ of B × B orthogonal dimensions conversionB×BIn randomly select MjIndividual row vector forms observing matrix Φj;
Utilize formulaWherein s ∈ { H, V, D }, 1≤l≤L, obtain observation set yj。
Specifically, according to formulaTry to achieve the adaptively sampled of every image block
Rate rj, and according to formula Mj=rj×B2Calculate each piece of adaptive observation number Mj, from the random of B × B orthogonal dimensions conversion
Matrix ΦB×BIn randomly select MjIndividual row vector forms observing matrix Φj, utilize observing matrix ΦjObtain observation set yjAs
Texture image data after processing.
Low frequency wavelet area image will be obtained after the high frequency coefficient zero setting of the small echo area image, by the low frequency wavelet domain figure
As obtaining estimating Flat image after wavelet inverse transformation;
The Flat image of estimating is subjected to the adjacent block edge adaptive weighted filter processing of two dimension, flat figure after being handled
As data, specifically include:
In described B pixel of symmetric extension of four direction up and down for estimating Flat image, estimated after being expanded
Flat image;
B for estimating Flat image after expansioni,jImage block, by Bi,jFour adjacent block B of image blocki-1,j、Bi+1,j、
Bi,j-1、Bi,j+1In, with Bi,jThe group of image block its adjacent a line, a row and four summits composition one (B+2) × (B+2)
Close block Bi,j', by combination block Bi,j' four summit zero setting;
To Bi,jB × B the pixel at ' center carries out smothing filtering, wherein:
To the pixel value b to be updated of the sub-block of the i-th row jth rowi,jSmoothly, obtain the sub-block renewal of the i-th row jth row
Pixel value b afterwardsi,j' be:
bi,j'=pc·bi,j+pi,j l·bi,0+pi,j r·bi,B+2+pi,j u·b0,j+pi,j d·bB+2,j, 1≤i, j≤B;
pc=0.5, p1=1/ (j)2,p2=1/ (B+1-j)2,p3=1/ (i)2,p4=1/ (B+1-i)2;
S=(p1+p2+p3+p4)·2;
pi,j l=p1/s,pi,j r=p2/s,pi,j u=p3/s,pi,j d=p4/ s, wherein:
pc、pi,j l、pi,j r、pi,j u、pi,j dFor five groups of smooth parameters, bi,0、bi,B+2、b0,j、bB+2,jFor pixel value to be updated
bi,jFour corresponding edge pixel values.
Specifically, as shown in figure 3, to Bi,jB × B the pixel at ' center carries out smothing filtering, it is assumed that to the i-th row jth
The sub-block of row is carried out smoothly, because the edge pixel gray scale between adjacent block produces blocking effect caused by identical saltus step, therefore
The pixel, which is filtered, need to utilize the local edge of four adjacent blocks around this pixel.The edge pixel values of adjacent block are entered
Weighted sum of the row based on distance, obtains the pixel value b after renewali,j′。
Texture image data after the processing and Flat image data after the processing are reconstructed, obtain reconstructing back
Texture image after edge image and reconstruct, texture image after the reconstruct back edge image and the reconstruct is overlapped to obtain weight
Build image.
Sixth embodiment of the invention provides a kind of storage medium, and the storage medium stores computer instruction, works as computer
When performing the computer instruction, for performing all steps of the multiple dimensioned splits' positions cognitive method of image adaptive as previously described
Suddenly.
Embodiment described above only expresses the several embodiments of the present invention, and its description is more specific and detailed, but simultaneously
Therefore the limitation to the scope of the claims of the present invention can not be interpreted as.It should be pointed out that for one of ordinary skill in the art
For, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to the guarantor of the present invention
Protect scope.Therefore, the protection domain of patent of the present invention should be determined by the appended claims.
Claims (11)
- A kind of 1. multiple dimensioned splits' positions cognitive method of image adaptive, it is characterised in that including:Small echo area image is obtained after image to be compressed is carried out into wavelet decomposition;High frequency wavelet area image will be obtained after the low frequency coefficient zero setting of the small echo area image, the high frequency wavelet area image is passed through Obtain estimating texture image after wavelet inverse transformation;By it is described estimate texture image carry out self-adaptive processing after, texture image data after being handled;Low frequency wavelet area image will be obtained after the high frequency coefficient zero setting of the small echo area image, the low frequency wavelet area image is passed through Obtain estimating Flat image after wavelet inverse transformation;By it is described estimate Flat image carry out self-adaptive processing after, Flat image data after being handled;Texture image data after the processing and Flat image data after the processing are reconstructed, obtain reconstructing back edge figure Texture image after picture and reconstruct, texture image after the reconstruct back edge image and the reconstruct is overlapped to obtain reconstruction figure Picture.
- 2. the multiple dimensioned splits' positions cognitive method of image adaptive according to claim 1, it is characterised in that:It is described by it is described estimate texture image carry out self-adaptive processing after, texture image data after being handled, specifically include:Will It is described estimate texture image carry out the adaptively sampled inspection process of entropy after, texture image data after being handled;It is described by it is described estimate Flat image carry out self-adaptive processing after, Flat image data, are specifically included after being handled:Will The Flat image of estimating carries out the adjacent block edge adaptive weighted filter processing of two dimension, Flat image data after being handled.
- 3. the multiple dimensioned splits' positions cognitive method of image adaptive according to claim 2, it is characterised in that described by institute State after estimating the texture image progress adaptively sampled inspection process of entropy, texture image data after being handled, specifically include:The texture image of estimating is divided into the image block that h size is B × B and non-overlapping copies;According to formulaTry to achieve the adaptive sample rate r of every image blockj, wherein, S is Default target sampling rate, h are piecemeal number, SminFor Least sampling rate threshold value, HjFor the gray level entropy of each image block;According to formula Mj=rj×B2Calculate j-th piece of adaptive observation number Mj;From the random matrix Φ of B × B orthogonal dimensions conversionB×BIn randomly select MjIndividual row vector forms observing matrix Φj;Utilize observing matrix ΦjObtain observation set yjAs texture image data after processing.
- 4. the multiple dimensioned splits' positions cognitive method of image adaptive according to claim 2, it is characterised in that described by institute State and estimate the adjacent block edge adaptive weighted filter processing of Flat image progress two dimension, Flat image data after being handled, specifically Including:In described B pixel of symmetric extension of four direction up and down for estimating Flat image, estimated after being expanded flat Image;B for estimating Flat image after expansioni,jImage block, by Bi,jFour adjacent block B of image blocki-1,j、Bi+1,j、Bi,j-1、 Bi,j+1In, with Bi,jThe combination block of image block its adjacent a line, a row and four summits composition one (B+2) × (B+2) Bi,j', by combination block Bi,j' four summit zero setting;To Bi,jB × B the pixel at ' center carries out smothing filtering, wherein:To the pixel value b to be updated of the sub-block of the i-th row jth rowi,jCarry out after smoothly, obtaining the sub-block renewal that the i-th row jth arranges Pixel value bi,j' be:bi,j'=pc·bi,j+pi,j l·bi,0+pi,j r·bi,B+2+pi,j u·b0,j+pi,j d·bB+2,j, 1≤i, j≤B;pc=0.5, p1=1/ (j)2,p2=1/ (B+1-j)2,p3=1/ (i)2,p4=1/ (B+1-i)2;S=(p1+p2+p3+p4)·2;pi,j l=p1/s,pi,j r=p2/s,pi,j u=p3/s,pi,j d=p4/ s, wherein:pc、pi,j l、pi,j r、pi,j u、pi,j dFor five groups of smooth parameters, bi,0、bi,B+2、b0,j、bB+2,jFor pixel value b to be updatedi,jInstitute Corresponding four edge pixel values.
- 5. according to any one of Claims 1 to 4 described image self-adapting multi-dimension splits' positions cognitive method, it is characterised in that Small echo area image is obtained after the progress wavelet decomposition by image to be compressed, is specifically included:Image to be compressed is carried out three layers small Small echo area image is obtained after Wave Decomposition.
- 6. a kind of electronic equipment, it is characterised in that including:At least one processor;AndThe memory being connected with least one processor communication;Wherein,The memory storage has can be by the instruction of one computing device, and the instruction is by least one processor Perform, so that at least one processor can:Small echo area image is obtained after image to be compressed is carried out into wavelet decomposition;High frequency wavelet area image will be obtained after the low frequency coefficient zero setting of the small echo area image, the high frequency wavelet area image is passed through Obtain estimating texture image after wavelet inverse transformation;By it is described estimate texture image carry out self-adaptive processing after, texture image data after being handled;Low frequency wavelet area image will be obtained after the high frequency coefficient zero setting of the small echo area image, the low frequency wavelet area image is passed through Obtain estimating Flat image after wavelet inverse transformation;By it is described estimate Flat image carry out self-adaptive processing after, Flat image data after being handled;Texture image data after the processing and Flat image data after the processing are reconstructed, obtain reconstructing back edge figure Texture image after picture and reconstruct, texture image after the reconstruct back edge image and the reconstruct is overlapped to obtain reconstruction figure Picture.
- 7. electronic equipment according to claim 6, it is characterised in that:It is described by it is described estimate texture image carry out self-adaptive processing after, texture image data after being handled, specifically include:Will It is described estimate texture image carry out the adaptively sampled inspection process of entropy after, texture image data after being handled;It is described by it is described estimate Flat image carry out self-adaptive processing after, Flat image data, are specifically included after being handled:Will The Flat image of estimating carries out the adjacent block edge adaptive weighted filter processing of two dimension, Flat image data after being handled.
- 8. electronic equipment according to claim 7, it is characterised in that it is described by it is described estimate texture image carry out entropy it is adaptive After inspection process should being sampled, texture image data after being handled, specifically include:The texture image of estimating is divided into the image block that h size is B × B and non-overlapping copies;According to formulaTry to achieve the adaptive sample rate r of every image blockj, wherein, S is Default target sampling rate, h are piecemeal number, SminFor Least sampling rate threshold value, HjFor the gray level entropy of each image block;According to formula Mj=rj×B2Calculate j-th piece of adaptive observation number Mj;From the random matrix Φ of B × B orthogonal dimensions conversionB×BIn randomly select MjIndividual row vector forms observing matrix Φj;Utilize observing matrix ΦjObtain observation set yjAs texture image data after processing.
- 9. electronic equipment according to claim 7, it is characterised in that described that the Flat image of estimating is subjected to two dimension neighbour The processing of block edge adaptive weighted filter, Flat image data, are specifically included after being handled:In described B pixel of symmetric extension of four direction up and down for estimating Flat image, estimated after being expanded flat Image;B for estimating Flat image after expansioni,jImage block, by Bi,jFour adjacent block B of image blocki-1,j、Bi+1,j、Bi,j-1、 Bi,j+1In, with Bi,jThe combination block of image block its adjacent a line, a row and four summits composition one (B+2) × (B+2) Bi,j', by combination block Bi,j' four summit zero setting;To Bi,jB × B the pixel at ' center carries out smothing filtering, wherein:To the pixel value b to be updated of the sub-block of the i-th row jth rowi,jCarry out after smoothly, obtaining the sub-block renewal that the i-th row jth arranges Pixel value bi,j' be:bi,j'=pc·bi,j+pi,j l·bi,0+pi,j r·bi,B+2+pi,j u·b0,j+pi,j d·bB+2,j, 1≤i, j≤B;pc=0.5, p1=1/ (j)2,p2=1/ (B+1-j)2,p3=1/ (i)2,p4=1/ (B+1-i)2;S=(p1+p2+p3+p4)·2;pi,j l=p1/s,pi,j r=p2/s,pi,j u=p3/s,pi,j d=p4/ s, wherein:pc、pi,j l、pi,j r、pi,j u、pi,j dFor five groups of smooth parameters, bi,0、bi,B+2、b0,j、bB+2,jFor pixel value b to be updatedi,jInstitute Corresponding four edge pixel values.
- 10. according to any one of claim 6~9 electronic equipment, it is characterised in that described that image to be compressed is carried out into small echo Small echo area image is obtained after decomposition, is specifically included:Small echo area image is obtained after image to be compressed is carried out into three layers of wavelet decomposition.
- 11. a kind of storage medium, it is characterised in that the storage medium stores computer instruction, when computer performs the meter When calculation machine instructs, for performing such as any one of Claims 1 to 5 described image self-adapting multi-dimension splits' positions cognitive method All steps.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710621894.XA CN107392975B (en) | 2017-07-27 | 2017-07-27 | Image self-adaptive multi-scale block compression sensing method and electronic equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710621894.XA CN107392975B (en) | 2017-07-27 | 2017-07-27 | Image self-adaptive multi-scale block compression sensing method and electronic equipment |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107392975A true CN107392975A (en) | 2017-11-24 |
CN107392975B CN107392975B (en) | 2021-02-09 |
Family
ID=60341761
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710621894.XA Active CN107392975B (en) | 2017-07-27 | 2017-07-27 | Image self-adaptive multi-scale block compression sensing method and electronic equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107392975B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109982090A (en) * | 2019-03-21 | 2019-07-05 | 西安电子科技大学 | A kind of adaptive splits' positions cognitive method of sample rate of combination gray level entropy and blind deconvolution |
CN110175965A (en) * | 2019-05-30 | 2019-08-27 | 齐齐哈尔大学 | Based on splits' positions cognitive method that is adaptively sampled and smoothly projecting |
CN110401839A (en) * | 2019-08-22 | 2019-11-01 | 上海无线电设备研究所 | A kind of space remote sensing method for compressing image based on the perception of adaptive splits' positions |
CN110717875A (en) * | 2019-10-18 | 2020-01-21 | 华北理工大学 | High-definition image processing method |
CN110827222A (en) * | 2019-11-04 | 2020-02-21 | 大连大学 | Wavelet domain image texture equalization method based on wave-type matrix permutation |
CN114666584A (en) * | 2022-04-02 | 2022-06-24 | 中山大学 | Coding method, device, equipment and medium based on high-frequency zero setting |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103559696A (en) * | 2013-09-29 | 2014-02-05 | 南京邮电大学 | Image fusion method based on compressed sensing |
CN104574450A (en) * | 2014-12-31 | 2015-04-29 | 南京邮电大学 | Image reconstruction method based on compressed sensing |
US20160178720A1 (en) * | 2014-12-17 | 2016-06-23 | Siemens Aktiengesellschaft | Memory Efficiency Of Parallel Magnetic Resonance Imaging Reconstruction |
CN105761290A (en) * | 2016-03-01 | 2016-07-13 | 北京理工大学 | Adaptive multi-scale partitioning compression sensing sampling method |
-
2017
- 2017-07-27 CN CN201710621894.XA patent/CN107392975B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103559696A (en) * | 2013-09-29 | 2014-02-05 | 南京邮电大学 | Image fusion method based on compressed sensing |
US20160178720A1 (en) * | 2014-12-17 | 2016-06-23 | Siemens Aktiengesellschaft | Memory Efficiency Of Parallel Magnetic Resonance Imaging Reconstruction |
CN104574450A (en) * | 2014-12-31 | 2015-04-29 | 南京邮电大学 | Image reconstruction method based on compressed sensing |
CN105761290A (en) * | 2016-03-01 | 2016-07-13 | 北京理工大学 | Adaptive multi-scale partitioning compression sensing sampling method |
Non-Patent Citations (4)
Title |
---|
JAMES E.FOWLER ET AL: ""Multiscale block compressed sensing with smoothed projected Landweber reconstruction"", 《2011 19TH EUROPEAN SIGNAL PROCESSING CONFERENCE》 * |
孙丽卿: ""基于分块压缩传感的图像重建算法研究"", 《中国优秀硕士学位论文全文数据库-信息科技辑》 * |
张学全等: ""基于自适应空域滤波的图像去块效应算法"", 《计算机工程》 * |
李玉 等: ""边缘和方向估计的自适应多尺度分块压缩感知算法"", 《信号处理》 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109982090A (en) * | 2019-03-21 | 2019-07-05 | 西安电子科技大学 | A kind of adaptive splits' positions cognitive method of sample rate of combination gray level entropy and blind deconvolution |
CN109982090B (en) * | 2019-03-21 | 2020-07-14 | 西安电子科技大学 | Sampling rate self-adaptive block compression sensing method combining gray level entropy and blind deconvolution |
CN110175965A (en) * | 2019-05-30 | 2019-08-27 | 齐齐哈尔大学 | Based on splits' positions cognitive method that is adaptively sampled and smoothly projecting |
CN110401839A (en) * | 2019-08-22 | 2019-11-01 | 上海无线电设备研究所 | A kind of space remote sensing method for compressing image based on the perception of adaptive splits' positions |
CN110717875A (en) * | 2019-10-18 | 2020-01-21 | 华北理工大学 | High-definition image processing method |
CN110717875B (en) * | 2019-10-18 | 2023-08-08 | 华北理工大学 | High-definition image processing method |
CN110827222A (en) * | 2019-11-04 | 2020-02-21 | 大连大学 | Wavelet domain image texture equalization method based on wave-type matrix permutation |
CN114666584A (en) * | 2022-04-02 | 2022-06-24 | 中山大学 | Coding method, device, equipment and medium based on high-frequency zero setting |
Also Published As
Publication number | Publication date |
---|---|
CN107392975B (en) | 2021-02-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107392975A (en) | The multiple dimensioned splits' positions cognitive method of image adaptive, electronic equipment | |
CN110136066B (en) | Video-oriented super-resolution method, device, equipment and storage medium | |
CN111105352B (en) | Super-resolution image reconstruction method, system, computer equipment and storage medium | |
Yu et al. | A unified learning framework for single image super-resolution | |
US10311547B2 (en) | Image upscaling system, training method thereof, and image upscaling method | |
CN111179177A (en) | Image reconstruction model training method, image reconstruction method, device and medium | |
CN110136057B (en) | Image super-resolution reconstruction method and device and electronic equipment | |
JP2015225665A (en) | Image noise removal method and image noise removal device | |
JP2013518336A (en) | Method and system for generating an output image with increased pixel resolution from an input image | |
Suetake et al. | Image super-resolution based on local self-similarity | |
CN110288530A (en) | A kind of pair of image carries out the processing method and processing device of super-resolution rebuilding | |
Kim et al. | Pynet-ca: enhanced pynet with channel attention for end-to-end mobile image signal processing | |
Prathap et al. | Robust and blind watermarking scheme for three dimensional anaglyph images | |
Canh et al. | Multi-scale deep compressive imaging | |
Janjušević et al. | CDLNet: Noise-adaptive convolutional dictionary learning network for blind denoising and demosaicing | |
CN115841420A (en) | Polarization image super-resolution reconstruction method based on deep learning | |
CN111353982B (en) | Depth camera image sequence screening method and device | |
CN110930314B (en) | Image stripe noise suppression method and device, electronic equipment and storage medium | |
CN115797194A (en) | Image denoising method, image denoising device, electronic device, storage medium, and program product | |
CN113362338A (en) | Rail segmentation method, device, computer equipment and rail segmentation processing system | |
CN108010000B (en) | Compressed image sharpening method based on adjacent feature subspace fitting | |
Sun et al. | A novel multi-focus image fusion algorithm based on NSST-FRFT | |
CN112581362A (en) | Image processing method and device for adjusting image details | |
Zhang et al. | Reducing blocking artifacts in JPEG-compressed images using an adaptive neural network-based algorithm | |
Zha et al. | Conditional invertible image re-scaling |
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 |