CN108510453A - The intelligent traffic monitoring image deblurring method of view-based access control model attention mechanism - Google Patents
The intelligent traffic monitoring image deblurring method of view-based access control model attention mechanism Download PDFInfo
- Publication number
- CN108510453A CN108510453A CN201810188142.3A CN201810188142A CN108510453A CN 108510453 A CN108510453 A CN 108510453A CN 201810188142 A CN201810188142 A CN 201810188142A CN 108510453 A CN108510453 A CN 108510453A
- Authority
- CN
- China
- Prior art keywords
- image
- notable
- traffic monitoring
- deblurring
- value
- 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
- 238000012544 monitoring process Methods 0.000 title claims abstract description 43
- 238000000034 method Methods 0.000 title claims abstract description 31
- 230000007246 mechanism Effects 0.000 title claims abstract description 18
- 230000006870 function Effects 0.000 claims abstract description 21
- 238000012545 processing Methods 0.000 claims abstract description 20
- 230000011218 segmentation Effects 0.000 claims abstract description 17
- 238000003709 image segmentation Methods 0.000 claims abstract description 15
- 239000011159 matrix material Substances 0.000 claims description 30
- 230000008569 process Effects 0.000 claims description 7
- 238000012876 topography Methods 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 238000012880 independent component analysis Methods 0.000 claims description 4
- 238000013139 quantization Methods 0.000 claims description 4
- 230000007480 spreading Effects 0.000 claims description 4
- 230000003044 adaptive effect Effects 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 3
- 230000004438 eyesight Effects 0.000 description 8
- 230000000007 visual effect Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 210000004556 brain Anatomy 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000016776 visual perception Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/73—Deblurring; Sharpening
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/12—Edge-based segmentation
-
- 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
- G06T2207/10016—Video; Image sequence
-
- 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
- G06T2207/10024—Color image
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20021—Dividing image into blocks, subimages or windows
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30232—Surveillance
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of intelligent traffic monitoring image deblurring method of view-based access control model attention mechanism, step includes：The original traffic monitoring image that one width obscures is transformed into HSI color spaces by step 1, the notable figure for generating original traffic monitoring image by rgb color space；Then it according to the scene information of image, maximizes scene information and obtains notable figure；Step 2 carries out image segmentation using the contour feature and textural characteristics of notable figure, obtains the segmentation figure of notable figure；Step 3 carries out deblurring processing to segmentation image, carries out deblurring processing to the segmentation image of notable figure using structural information spread function, finally obtains the clear image after a secondary deblurring.The method of the present invention, step is simple, and committed memory space is few, the significant effect after deblurring.
Description
Technical field
The invention belongs to image deblurring processing technology fields, are related to a kind of intelligent transportation prison of view-based access control model attention mechanism
Control image deblurring method.
Background technology
As traffic monitoring and the intelligent level of traffic administration are continuously improved, with traffic monitoring image procossing, analysis, reason
Intelligent Video Surveillance Technology based on solution attracts people's attention more and more.However, traffic monitoring image is in practical bat
It takes the photograph, during transimission and storage etc., can all be influenced by factors such as imaging device, environment, noises, cause image fuzzy,
In, most commonly traffic monitoring camera exposure when due to camera and shooting object between relative motion caused by image transport
Dynamic model is pasted, and shooting object causes image defocus fuzzy with the improper of camera photocentre distance, these, which are obscured, can cause to hand over
Logical monitoring image material particular information is lost, and seriously affects traffic monitoring and intellectual management is horizontal.
With the raising of global auto recoverable amount sharply increased with people's awareness of safety, intelligent transportation supervisory systems without when
It is not playing an important role without carving, be used to ensure road safety passage and preventing emergency situations.Traffic monitoring image
Processing has very important effect in intelligent transportation supervisory systems, however, as image data scale becomes increasingly huger
Greatly, gradually become for the screening of magnanimity traffic monitoring image information and processing while the mankind can obtain unprecedented affluent resources
More difficult, traditional image deblurring processing method has been difficult to reach ideal effect.Therefore, how rapidly to screen and go
It urgently studies and solves the problems, such as except the fuzzy traffic monitoring image for providing high quality of image becomes.
Perception terminal of the human eye as visual image information, the high-level vision information processing system formed by long-term evolution
System, can efficiently, accurately handle the image information of input, have born selective power to the visual information of input, can be thousand
Become in the scene that ten thousand change and quickly and accurately judge, and sight is focused on interested important information, and then is subject to
Careful analysis and deciphering, human visual system can effectively complete this process and depend on vision noticing mechanism, this
Vision noticing mechanism of the kind comprising selectivity and initiative contributes to the various information in brain parallel processing field of vision, vision aobvious
Work property effectively assists vision noticing mechanism to realize the automatic real-time selection to target as vision noticing mechanism important content.
Therefore, studying intelligent traffic monitoring image deblurring method by means of human visual attention mechanism and improve picture quality has
Important practical significance.
Invention content
The object of the present invention is to provide a kind of intelligent traffic monitoring image deblurring method of view-based access control model attention mechanism, solutions
It has determined and has been directed to the defect of magnanimity traffic monitoring image information screened with processing difficulty in the prior art, it is difficult to quickly screen and go
Except the fuzzy traffic monitoring image for providing high quality of image, the problem for causing the intelligent level of traffic monitoring and management low.
The technical solution adopted in the present invention is a kind of intelligent traffic monitoring image deblurring of view-based access control model attention mechanism
Method is implemented according to the following steps：
Step 1, the notable figure for generating original traffic monitoring image,
The original traffic monitoring image that one width obscures is transformed into HSI color spaces by rgb color space；Then according to figure
The scene information of picture maximizes scene information and obtains notable figure；
Step 2 carries out image segmentation using the contour feature and textural characteristics of notable figure, obtains the segmentation figure of notable figure；
Step 3 carries out deblurring processing to segmentation image,
Deblurring processing is carried out to the segmentation image of notable figure using structural information spread function, a pair is finally obtained and removes mould
Clear image after paste.
The invention has the advantages that acquiring the notable figure of original traffic monitoring image, profit using maximization scene information
Image segmentation is carried out with the contour feature of the notable figure and textural characteristics, is carried out at deblurring using structural information spread function
Reason finally exports the image after deblurring.The present invention has method is simple, is removed traffic by means of human visual attention mechanism to supervise
The advantages that image is fuzzy is controlled, can be used for handling generally fuzzy coloured image.
Description of the drawings
Fig. 1 is the flow chart of the method for the present invention；
Fig. 2 a are the image before fuzzy using present invention removal traffic monitoring image；
Fig. 2 b are the notable figure before fuzzy using present invention removal traffic monitoring image；
Fig. 3 a are to carry out the segmentation figure that image segmentation obtains using the present invention；
Fig. 3 b are to carry out the edge graph that image segmentation obtains using the present invention；
Fig. 3 c are to carry out the energy diagram that image segmentation obtains using the present invention；
Fig. 4 a are the image after fuzzy using present invention removal traffic monitoring image；
Fig. 4 b are the notable figure after fuzzy using present invention removal traffic monitoring image；
Fig. 5 a are to remove the image before general coloured image obscures using the present invention；
Fig. 5 b are to remove the notable figure before general coloured image obscures using the present invention；
Fig. 6 a are to carry out the segmentation figure that image segmentation obtains using the present invention；
Fig. 6 b are to carry out the edge graph that image segmentation obtains using the present invention；
Fig. 6 c are to carry out the energy diagram that image segmentation obtains using the present invention；
Fig. 7 a are to remove the image after general coloured image obscures using the present invention；
Fig. 7 b are to remove the notable figure after general coloured image obscures using the present invention；
Fig. 8 is that the image deblurring conversion main process substep of the embodiment of the present invention 1 is schemed.
Specific implementation mode
The following describes the present invention in detail with reference to the accompanying drawings and specific embodiments.
The present invention is based on the intelligent traffic monitoring image deblurring methods of vision noticing mechanism, implement according to the following steps：
Step 1, the notable figure for generating original traffic monitoring image, the original traffic monitoring image that a width is obscured is by RGB
Color space is transformed into the HSI color spaces for more meeting human visual perception system；Then maximum according to the scene information of image
Change scene information and obtain notable figure,
The calculating formula that original traffic monitoring image is converted to HSI color spaces by rgb color space is：
The notable figure that scene information acquires image is maximized, detailed process is as follows：
1.1) it is directed to original traffic monitoring image and base vector is calculated using independent component analysis method, extraction conversioning colour is empty
Between after image X internal feature,
The calculating formula of independent component analysis method is X=AS, wherein image X is mixed by independent element S with hybrid matrix A
It forms, X={ x1(t),x2(t),…,xn(t)}T, t is time samples, S={ s1,s2,…,sn}T, A=(m × n) is mixed moment
Battle array；Further calculate the inverse matrix A of matrix A-1, solve inverse matrix and the component of regional area picture element matrix be separated into isolated component；
Image is subjected to piecemeal processing again, regional area picture element matrix obtains the base of regional area picture element matrix after being multiplied with solution inverse matrix
Vectorial W={ w1,w2,…,wn}；
1.2) estimate to carry out possibility predication to each component of base vector W using Gaussian Kernel Density, for regional area pixel
The calculating formula of block is as follows：
In calculating formula (1), p functions refer to the Gaussian Kernel Density estimated value of topography's block,
Wherein, σ is scale factor, and regional area pixel block size is j × k, and value is 5 × 5 herein；ψ is indicated at whole picture
The image of reason, due to each component w of vectorial WiBetween independently of each other, wherein component wiValue be vi；ω (s, t) is cuclear density letter
The weights of Gaussian function in number, for the weights by the base system number probability Estimation of current topography, calculating formula is as follows：
1.3) notable figure is measured by calculating regional area block of pixels self-information, regional area block of pixels self-information amount
Calculating formula is as follows：
Wherein, I (x) is regional area block of pixels self-information amount, and p (x) is the Gaussian kernel of topography's block in step 1.2)
Density estimation value；
Step 2 carries out image segmentation using the contour feature and textural characteristics of notable figure, obtains the segmentation figure of notable figure,
It is as follows：
2.1) it is directed to the notable figure extracted in step 1, randomly selects K object as cluster initial point v1,v2,
...vk, the distance between each object and each cluster centre are calculated, the nearest person of chosen distance gives in the cluster nearest from it
The heart distributes whole objects, and each cluster centre then constantly computes repeatedly cluster according to existing object, and calculating formula is：
Wherein, xi(ki+ 1) it is i-th of object, kiIndicate the clustering object sum where i-th of object；
2.2) recursive operation is carried out using K mean cluster method, entire image is divided into small and continuous k image block
Region；
2.3) image block areas obtained in step 2.2) is changed into undirected weighted graph Gc=(Vc,Ec；Wc), then count
The weight matrix W of image block areas is calculated, calculating formula is as follows：
Wherein, WijIndicate the weights between undirected weighted graph G interior joints i and node j, indicate in the picture region i and j it
Between relationship；F (i) and F (j) indicates the i-th region v respectivelyiWith j-th of region vjGray value；σIAnd σVFor adjustment parameter；||V
(i)-V(j)||2For manhatton distance, r is data and the distance of barycenter, is obtained by adaptive polo placement；
2.4) characteristic equation (D is calculatedc-Wc)yc=λ Dcyc,In characteristic value and feature vector,
Wherein, Dc-WcFor Laplacian Matrix, Wc(i, j)=wc(i,j),Dc(i, j)=∑jwc(i,j)；
ycFor instruction vector, y is indicatedcIn one region of each element representation；
2.5) divide image, two parts are divided the image into using the second minimal characteristic vector in step 2.4), by Yc
In be more than 0 region element be divided into a group, by YcIn be less than 0 region element be divided into another group；
2.6) image after recursive call step 2.4), step 2.5) are divided.
Step 3 carries out deblurring processing to segmentation image,
Deblurring processing is carried out to the segmentation image of notable figure using structural information spread function, calculating formula is：
Wherein, i, j are respectively the position coordinates of image slices vegetarian refreshments, and truncated error is O (τ+h2), time discrete walks in above formula
Long τ preferred values are 5, and spatial spreading step-length h preferred values are 400, and iterations n preferred values are 10,
In addition, (Ix)i,j=(2 (Ii+1,j-Ii-1,j)+Ii+1,j+1-Ii-1,j+1+Ii+1,j-1-Ii-1,j-1)/4,
(Iy)i,j=(2 (Ii,j+1-Ii,j-1)+Ii+1,j+1-Ii+1,j-1+Ii-1,j+1-Ii-1,j-1)/4；
GαFor gaussian kernel function, calculating formula is：
Wherein, α is scale parameter, and preferably value is 1；
The calculating formula of spread function g (| t |) is：
Wherein,Calculating formula is as follows：
Wherein, it is mean value, μ that the preferred value of quantization parameter k, which is 2, μ, in reaction item f (I)1Value is 13, v1Value is 45,
μ2Value is 68, v2Value is 125, μ3Value is 205；
SFijFor structural information function, calculating formula is as follows：
Wherein,For the gradient of any pixel point in image g,
By above-mentioned calculating, the clear image after a secondary deblurring is finally obtained,.
Following embodiment is all made of Matlab R2017a programmings and realizes method described in the invention.Experiment porch configures：
Operating system is Windows10, and CPU is Intel Core i7 5600U, RAM 8G.
Embodiment 1
It is as follows for removing traffic monitoring image and obscuring with reference to Fig. 8：
Step 1, the notable figure for generating original traffic monitoring image.
It is the fuzzy traffic monitoring original image of a width as shown in Figure 2, first, by the rgb color space of the original image
HSI color spaces are converted to, calculating formula is as follows：
Secondly, the internal feature of image X after convert color spaces, the inverse matrix A of matrix A in calculating formula X=AS are extracted-1,
The component of regional area picture element matrix is separated into isolated component by solution inverse matrix.Image is subjected to piecemeal processing, regional area picture
Prime matrix obtains the base vector W={ w of regional area picture element matrix after being multiplied with solution inverse matrix1,w2,…,wn}.Utilize Gaussian kernel
Density estimation carries out possibility predication to each component of base vector W, and calculating formula is：
Wherein, regional area pixel block size is j × k, and value is 5 × 5 herein；ψ indicates the image of whole picture processing, due to
Each component w of vectorial WiBetween independently of each other, wherein component wiValue be vi, ω (s, t) is Gaussian function in kernel density function
Weights, for the weights by the base system number probability Estimation of current topography, calculating formula is as follows：
Finally, it calculates regional area block of pixels self-information amount and obtains notable figure, calculating formula is：
Step 2 carries out image segmentation using the contour feature and textural characteristics of notable figure,
First, recursive operation is carried out using K mean cluster method, entire image is divided into small and continuous k image
Block region, calculation formula are：
Wherein, xi(ki+ 1) it is i-th of object, kiIndicate the clustering object sum where i-th of object；
Secondly, divide image, calculate the undirected weighted graph G of image block areas being partitioned into previous stepc=(Vc,Ec；Wc)
In weight matrix W, calculating formula is as follows：
Wherein, WijIndicate the weights between undirected weighted graph G interior joints i and node j, indicate in the picture region i and j it
Between relationship；F (i) and F (j) indicates the i-th region v respectivelyiWith j-th of region vjGray value, | | V (i)-V (j) | |2For graceful Kazakhstan
Distance, σIAnd σVFor adjustment parameter, respectively 58,128；
Calculate characteristic equation (Dc-Wc)yc=λ Dcyc,Characteristic value and feature vector,
Wherein, Wc(i, j)=wc(i,j),Dc(i, j)=∑jwc(i,j)；ycFor instruction vector, y is indicatedcIn each
One region of element representation；Dc-WcFor Laplacian Matrix；Y is calculatedcIn be more than 0 region element be divided into a group, Yc
In be less than 0 region element be divided into another group carry out image segmentation.
Step 3 handles segmentation image deblurring, and calculating formula is：
I, j are respectively the position coordinates of image slices vegetarian refreshments；Truncated error is O (τ+h2), time discrete step-length τ preferred values are
5, spatial spreading step-length h preferred values are 400, and iterations n preferred values are 10,
In addition, (Ix)i,j=(2 (Ii+1,j-Ii-1,j)+Ii+1,j+1-Ii-1,j+1+Ii+1,j-1-Ii-1,j-1)/4,
(Iy)i,j=(2 (Ii,j+1-Ii,j-1)+Ii+1,j+1-Ii+1,j-1+Ii-1,j+1-Ii-1,j-1)/4,
GαFor gaussian kernel function, calculating formula is as follows：
Wherein, α is scale parameter, is 1 in this value, spread function g (| t |) it is obtained by following calculating formula：
Wherein,It is obtained by following calculating formula：
It is mean value, μ that quantization parameter k values, which are 2, μ, in above formula, in reaction item f (I)1Value is 13, μ2Value is 68, μ3It takes
Value is 205, v1Value is 45, v2Value is 125；
SFijFor structural information function, obtained by following calculating formula：
Wherein,For the gradient of any pixel point in image g, the image after removal obscures is obtained,
Image after output removal is fuzzy, as shown in Figure 4, the license plate number marginal information of automobile is prominent in Fig. 4 a images therein
Go out, picture quality significantly improves；In salient region ratio Fig. 2 b in Fig. 4 b images therein more significantly, visual effect compared with
It is good.
Embodiment 2
For fuzzy to the removal of general coloured image, it is as follows：
In the present embodiment, Fig. 5 is the image before fuzzy using present invention removal coloured image.Fuzzy coloured image calculates
Notable figure step 1 is same as Example 1, and the image before removal is fuzzy is shown in Fig. 5 a and Fig. 5 b respectively with its notable figure.
In step 2, parameter value and the implementation of image segmentation are carried out using the contour feature of notable figure and textural characteristics
Example 1 is identical, and segmentation figure, edge graph and the energy diagram that image segmentation obtains are shown in Fig. 6 a, Fig. 6 b and Fig. 6 c respectively.
In step 3, Fuzzy Processing is removed to segmentation image, the process in the step 3 is same as Example 1, removes mould
Image after paste is shown in that image edge information protrudes in Fig. 7 a and Fig. 7 b, Fig. 7 a with its notable figure result, and picture quality significantly improves,
In salient region ratio Fig. 5 b in Fig. 7 b more significantly, visual effect is preferable.
Claims (5)
1. a kind of intelligent traffic monitoring image deblurring method of view-based access control model attention mechanism, which is characterized in that according to following step
It is rapid to implement：
Step 1, the notable figure for generating original traffic monitoring image,
The original traffic monitoring image that one width obscures is transformed into HSI color spaces by rgb color space；Then according to image
Scene information maximizes scene information and obtains notable figure；
Step 2 carries out image segmentation using the contour feature and textural characteristics of notable figure, obtains the segmentation figure of notable figure；
Step 3 carries out deblurring processing to segmentation image,
Deblurring processing is carried out to the segmentation image of notable figure using structural information spread function, after finally obtaining a secondary deblurring
Clear image.
2. the intelligent traffic monitoring image deblurring method of view-based access control model attention mechanism according to claim 1, feature
It is, in the step 1, maximizes the notable figure that scene information acquires image, detailed process is as follows：
1.1) it is directed to original traffic monitoring image and base vector is calculated using independent component analysis method, after extracting convert color spaces
The internal feature of image X,
The calculating formula of independent component analysis method is X=AS, wherein image X is mixed by independent element S and hybrid matrix A,
X={ x1(t),x2(t),…,xn(t)}T, t is time samples, S={ s1,s2,…,sn}T, A=(m × n) is hybrid matrix；Into
The inverse matrix A of one step calculating matrix A-1, solve inverse matrix and the component of regional area picture element matrix be separated into isolated component；It again will figure
As carrying out piecemeal processing, regional area picture element matrix obtains the base vector W of regional area picture element matrix after being multiplied with solution inverse matrix
={ w1,w2,…,wn}；
1.2) estimate to carry out possibility predication to each component of base vector W using Gaussian Kernel Density, for regional area block of pixels
Calculating formula is as follows：
In calculating formula (1), p functions refer to the Gaussian Kernel Density estimated value of topography's block,
Wherein, σ is scale factor, and regional area pixel block size is j × k；ψ indicates the image of whole picture processing, due to vectorial W's
Each component wiBetween independently of each other, wherein component wiValue be vi；ω (s, t) is the weights of Gaussian function in kernel density function, should
For weights by the base system number probability Estimation of current topography, calculating formula is as follows：
1.3) notable figure, the calculating of regional area block of pixels self-information amount are measured by calculating regional area block of pixels self-information
Formula is as follows：
Wherein, I (x) is regional area block of pixels self-information amount, and p (x) is the Gaussian Kernel Density of topography's block in step 1.2)
Estimated value.
3. the intelligent traffic monitoring image deblurring method of view-based access control model attention mechanism according to claim 2, feature
It is, in the step 2, detailed process is：
2.1) it is directed to the notable figure extracted in step 1, randomly selects K object as cluster initial point v1,v2,...vk, meter
The distance between each object and each cluster centre are calculated, the nearest person of chosen distance gives the cluster centre nearest from it, distribution
Whole objects, each cluster centre then constantly compute repeatedly cluster according to existing object, and calculating formula is：
Wherein, xi(ki+ 1) it is i-th of object, kiIndicate the clustering object sum where i-th of object；
2.2) recursive operation is carried out using K mean cluster method, entire image is divided into small and continuous k image block area
Domain；
2.3) image block areas obtained in step 2.2) is changed into undirected weighted graph Gc=(Vc,Ec；Wc), then calculate image
The weight matrix W in block region, calculating formula are as follows：
Wherein, WijIt indicates the weights between undirected weighted graph G interior joints i and node j, indicates between region i and j in the picture
Relationship；F (i) and F (j) indicates the i-th region v respectivelyiWith j-th of region vjGray value；σIAnd σVFor adjustment parameter；||V(i)-
V(j)||2For manhatton distance, r is data and the distance of barycenter, is obtained by adaptive polo placement；
2.4) characteristic equation (D is calculatedc-Wc)yc=λ Dcyc, in characteristic value and feature vector,
Wherein, Dc-WcFor Laplacian Matrix, Wc(i, j)=wc(i,j),Dc(i, j)=∑jwc(i,j)；
ycFor instruction vector, y is indicatedcIn one region of each element representation；
2.5) divide image, two parts are divided the image into using the second minimal characteristic vector in step 2.4), by YcIn be more than
The element in 0 region is divided into a group, by YcIn be less than 0 region element be divided into another group；
2.6) image after recursive call step 2.4), step 2.5) are divided.
4. the intelligent traffic monitoring image deblurring method of view-based access control model attention mechanism according to claim 3, feature
It is, in the step 3, detailed process is：
Deblurring processing calculating formula be：
Wherein, i, j are respectively the position coordinates of image slices vegetarian refreshments, and truncated error is O (τ+h2), time discrete step-length is in above formula
τ, spatial spreading step-length be h, iterations n,
In addition, (Ix)i,j=(2 (Ii+1,j-Ii-1,j)+Ii+1,j+1-Ii-1,j+1+Ii+1,j-1-Ii-1,j-1)/4,
(Iy)i,j=(2 (Ii,j+1-Ii,j-1)+Ii+1,j+1-Ii+1,j-1+Ii-1,j+1-Ii-1,j-1)/4；
GαFor gaussian kernel function, calculating formula is：
Wherein, α is scale parameter；
The calculating formula of spread function g (| t |) is：
Wherein,Calculating formula is as follows：
Wherein, k is quantization parameter in reaction item f (I), and μ is mean value, μ1、μ2、μ3、v1、v2It is parameter；
SFijFor structural information function, calculating formula is as follows：
Wherein,For the gradient of any pixel point in image g.
5. the intelligent traffic monitoring image deblurring method of view-based access control model attention mechanism according to claim 4, feature
It is, the time discrete step-length τ values are 5, and spatial spreading step-length h values are 400, and iterations n values are 10；
Scale parameter α values are 1, and quantization parameter k values are 2, μ1Value is 13, μ2Value is 68, μ3Value is 205, v1Value
It is 45, v2Value is 125.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810188142.3A CN108510453B (en) | 2018-03-07 | 2018-03-07 | Intelligent traffic monitoring image deblurring method based on visual attention mechanism |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810188142.3A CN108510453B (en) | 2018-03-07 | 2018-03-07 | Intelligent traffic monitoring image deblurring method based on visual attention mechanism |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108510453A true CN108510453A (en) | 2018-09-07 |
| CN108510453B CN108510453B (en) | 2021-06-29 |
Family
ID=63377096
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810188142.3A Expired - Fee Related CN108510453B (en) | 2018-03-07 | 2018-03-07 | Intelligent traffic monitoring image deblurring method based on visual attention mechanism |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108510453B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110472086A (en) * | 2019-08-02 | 2019-11-19 | 西安工程大学 | A kind of skeleton image search method extracted based on retina key feature |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110142370A1 (en) * | 2009-12-10 | 2011-06-16 | Microsoft Corporation | Generating a composite image from video frames |
| CN103514582A (en) * | 2012-06-27 | 2014-01-15 | 郑州大学 | Visual saliency-based image deblurring method |
| CN103884431A (en) * | 2013-12-31 | 2014-06-25 | 华中科技大学 | Infrared imaging detection and positioning method of underground building in plane surface environment |
| CN104616248A (en) * | 2014-11-20 | 2015-05-13 | 杭州电子科技大学 | Single image deblurring method combined with margin analysis and total variation |
| CN106097256A (en) * | 2016-05-31 | 2016-11-09 | 南京邮电大学 | A kind of video image fuzziness detection method based on Image Blind deblurring |
| CN107194927A (en) * | 2017-06-13 | 2017-09-22 | 天津大学 | The measuring method of stereo-picture comfort level chromaticity range based on salient region |
| CN107240119A (en) * | 2017-04-19 | 2017-10-10 | 北京航空航天大学 | Utilize the method for improving the fuzzy clustering algorithm extraction uneven infrared pedestrian of gray scale |
-
2018
- 2018-03-07 CN CN201810188142.3A patent/CN108510453B/en not_active Expired - Fee Related
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110142370A1 (en) * | 2009-12-10 | 2011-06-16 | Microsoft Corporation | Generating a composite image from video frames |
| CN103514582A (en) * | 2012-06-27 | 2014-01-15 | 郑州大学 | Visual saliency-based image deblurring method |
| CN103884431A (en) * | 2013-12-31 | 2014-06-25 | 华中科技大学 | Infrared imaging detection and positioning method of underground building in plane surface environment |
| CN104616248A (en) * | 2014-11-20 | 2015-05-13 | 杭州电子科技大学 | Single image deblurring method combined with margin analysis and total variation |
| CN106097256A (en) * | 2016-05-31 | 2016-11-09 | 南京邮电大学 | A kind of video image fuzziness detection method based on Image Blind deblurring |
| CN107240119A (en) * | 2017-04-19 | 2017-10-10 | 北京航空航天大学 | Utilize the method for improving the fuzzy clustering algorithm extraction uneven infrared pedestrian of gray scale |
| CN107194927A (en) * | 2017-06-13 | 2017-09-22 | 天津大学 | The measuring method of stereo-picture comfort level chromaticity range based on salient region |
Non-Patent Citations (5)
| Title |
|---|
| BING ZHOU 等: "Image Deblurring Based On Visual Saliency", 《2012 INTERNATIONAL CONFERENCE ON SYSTEMS AND INFORMATICS (ICSAI 2012)》 * |
| 吴晓旭等: "局部加权全变差下的盲去模糊", 《计算机辅助设计与图形学学报》 * |
| 王维哲等: "基于视觉显著性的运动图像去模糊研究", 《计算机工程与设计》 * |
| 郑杨帆: "基于图论的谱聚类算法研究和实现", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 韩鹏: "视觉注意力计算模型的研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110472086A (en) * | 2019-08-02 | 2019-11-19 | 西安工程大学 | A kind of skeleton image search method extracted based on retina key feature |
| CN110472086B (en) * | 2019-08-02 | 2023-01-31 | 西安工程大学 | Skeleton image retrieval method based on retina key feature extraction |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108510453B (en) | 2021-06-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109753903B (en) | Unmanned aerial vehicle detection method based on deep learning | |
| Bahrami et al. | A fast approach for no-reference image sharpness assessment based on maximum local variation | |
| Li | On incremental and robust subspace learning | |
| JP6884517B2 (en) | Information processing equipment, information processing methods and programs | |
| CN112767711B (en) | Multi-class multi-scale multi-target snapshot method and system | |
| CN110490252B (en) | Indoor people number detection method and system based on deep learning | |
| CN109918971B (en) | Method and device for detecting number of people in monitoring video | |
| CN111861925A (en) | Image rain removing method based on attention mechanism and gate control circulation unit | |
| CN111783589B (en) | Complex scene crowd counting method based on scene classification and multi-scale feature fusion | |
| CN109766828A (en) | A kind of vehicle target dividing method, device and communication equipment | |
| Marana et al. | Real-time crowd density estimation using images | |
| CN111241924A (en) | Face detection and alignment method and device based on scale estimation and storage medium | |
| CN110533632A (en) | Image obscures altering detecting method, device, computer equipment and storage medium | |
| CN109961425A (en) | A kind of water quality recognition methods of Dynamic Water | |
| CN111091093A (en) | Method, system and related device for estimating number of high-density crowds | |
| CN111986163A (en) | Face image selection method and device | |
| CN111753671A (en) | Crowd counting method for real scene | |
| CN115861799A (en) | Light-weight air-to-ground target detection method based on attention gradient | |
| CN111683221A (en) | Real-time video monitoring method and system for natural resources embedded with vector red line data | |
| Wei et al. | Sidgan: Single image dehazing without paired supervision | |
| CN109271864B (en) | Crowd density estimation method based on wavelet transformation and support vector machine | |
| CN108510453A (en) | The intelligent traffic monitoring image deblurring method of view-based access control model attention mechanism | |
| CN108769543A (en) | The determination method and device of time for exposure | |
| CN112907972A (en) | Road vehicle flow detection method and system based on unmanned aerial vehicle and computer readable storage medium | |
| CN109657083B (en) | Method and device for establishing textile picture feature library |
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: 20210629 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |