CN104966279A - Image synthesis restoration method based on local structure features - Google Patents

Abstract

The present invention relates to an image synthesis restoration method based on local structure features, comprising the following steps: inputting an image to be restored; initializing confidence coefficients of all pixel points in the image to be restored; selecting a template block of each pixel point by respectively taking each pixel point on a damaged boundary of the image to be restored as the center; calculating the priority of the template block where each pixel point on the damaged boundary of the image to be restored locates, and performing descending sort; searching an area which has the highest similarity with the template block having the greatest priority from an undamaged area in the image to be restored to serve as an optimal matching block; copying pixel values in the optimal matching block to the template block; and updating the confidence coefficients and returning to execute corresponding steps until the image to be restored is completely restored. The image synthesis restoration method based on local structure features provided in the present invention can improve image restoration accuracy for a damaged image with a complicated texture structure. The present invention also relates to an image synthesis restoration system based on local structure features.

Description

A kind of Images uniting restorative procedure based on partial structurtes feature

Technical field

The present invention relates to image processing field, be specifically related to a kind of Images uniting restorative procedure based on partial structurtes feature.

Background technology

Image repair is one of main contents of image procossing, mainly comprise the reparation of structural images and the reparation of texture image, restorative procedure is roughly divided into two classes: a class is the image repair method based on partial differential equation, and an other class is the image repair method based on textures synthesis.

Based in the image repair method of textures synthesis, two classes can be divided into again: a kind of be based on decompose method, another is the textures synthesis restorative procedure based on sample, and based in the textures synthesis restorative procedure of sample, it represents algorithm is the Criminisi algorithm driven based on isophote, Criminisi algorithm is the image block first determining to obtain at first reparation by priority formula, then searches for optimal objective block to fill defect area according to color-match criterion in global scope.But along with the carrying out of repair process, the data item of priority level initializing can decline rapidly, and degree of confidence item then starts to increase, and makes the calculating of priority unreliable, the fill order led to errors; And the global search mode adopted, time complexity can be higher; In addition, matching criterior just adopts color distance to determine blocks and optimal matching blocks, finally can produce multiple object block, and be the modes adopting random selecting object block for these multiple object block Criminisi algorithms, so just increase the erroneous matching rate of image.Finally show that the defect on image is exactly picture structure fidelity defect, image repair is of poor quality.

Summary of the invention

The invention provides a kind of Images uniting restorative procedure based on partial structurtes feature, image repair accuracy rate can be improved to the breakage image of texture structure more complicated.

The technical scheme that the present invention solves the problems of the technologies described above is as follows: a kind of Images uniting restorative procedure based on partial structurtes feature, comprises the following steps:

Step 1, input an image to be repaired, be designated as u (x, y), x, y are respectively the coordinate of image to be repaired on x, y direction;

Step 2, to the degree of confidence initialization of all pixels in described image to be repaired; Specific as follows:

$C\left(p^{\′}\right)=0,\∀p^{\′}\∈\mathrm{\Ω},C\left(p^{\′}\right)=1,\∀p^{\′}\∈\mathrm{\Φ}$

In formula, p' is the pixel in described image to be repaired; The degree of confidence that C (p') is pixel p'; Ω is the affected area in described image to be repaired, and Φ is the non-affected area in described image to be repaired; for getting arbitrary value; refer to that the affected area in described image to be repaired gets any pixel; refer to that the non-affected area in described image to be repaired gets any pixel;

Step 3, chooses the formwork of each pixel described respectively in described image damaged boundary to be repaired centered by each pixel;

Step 4, according to the size of the damaged degree priority of each pixel place formwork in image damaged boundary to be repaired described in following formulae discovery, and by damaged degree priority according to descending sort from big to small;

P(p)＝a ₁C(p)+a ₂D(p)+a ₃H(p)；

Wherein, p is the pixel of described formwork center; The damaged degree priority that P (p) is pixel p; a ₁, a ₂, a ₃for weight, and a ₁, a ₂, a ₃and be 1;

H (p)=Kh+exp (-h); K is controling parameter, and value is 0.8; Exp (-h)=e ^-h, e ^-hfor taking e as the exponential function of the truth of a matter, e=2.71828; H is local metric function, h=| λ ₁-λ ₂| ², λ ₁and λ ₂be respectively the First Eigenvalue and the Second Eigenvalue of described formwork structure tensor expression formula, expression formula is:

${\mathrm{\λ}}_{1,2}=\frac{1}{2}(j_{11}+j_{22}\±\sqrt{{(j_{11}+j_{22})}^2+4j_{12}^2})$

Wherein, j ₁₁, j ₁₂, j ₂₂obtained by its structure tensor expression formula to be repaired, described its structure tensor expression formula to be repaired is:

refer to G _ρwith ask convolution algorithm; refer to G _ρwith ask convolution algorithm;

refer to G _ρwith ask convolution algorithm;

C (p) is the degree of confidence of described formwork center pixel p; D (p) is the data item of described formwork center pixel p;

$C\left(p\right)=\frac{\underset{\hat{p}\∈{\mathrm{\Ψ}}_P\∩\mathrm{\Φ}}{\Σ}C\left(\hat{p}\right)}{\left|{\mathrm{\Ψ}}_p\right|}$

$D\left(p\right)=\frac{|\▿I_p^{\⊥}\·n_p|}{\mathrm{\α}}$

In formula, ψ _pfor the formwork that damaged boundary is put centered by pixel p, Φ is the non-affected area in described image to be repaired, ψ _pthe common factor of the non-affected area in ∩ Φ finger print plate and described image to be repaired, for the pixel in the common factor of the non-affected area in described formwork and described image to be repaired, i.e. the pixel of non-affected area in described formwork, for pixel degree of confidence, | Ψ _p| be the area of formwork, the quantity of pixel in finger print plate; for the direction of the isophote of pixel p, i.e. the vertical direction of gradient, n _pfor the unit direction vector of pixel p, α is normalized factor, and in gray level image, value is 255; $C\left(\hat{p}\right)=C\left(p^{\′}\right)=1,\∀p^{\′}\∈\mathrm{\Φ};$

Step 5, finds the formwork maximum with described damaged degree priority in the non-affected area of described image to be repaired the region that similarity is the highest, as blocks and optimal matching blocks;

Step 6, copies to the formwork that described damaged degree priority is maximum by the pixel value in described blocks and optimal matching blocks in;

Step 7, upgrades the maximum formwork of described damaged degree priority according to following formula damaged area in the degree of confidence of all pixels, and return step 4, until image to be repaired is repaired completely;

$C\left(\hat{q}\right)=C\left(p_m\right),\∀\hat{q}\∈{\mathrm{\Ψ}}_{p_m}\∩\mathrm{\Ω};$

In formula, p _mfor the pixel of the center of the maximum formwork of described damaged degree priority; C (p _m) be pixel p _mdegree of confidence, for the formwork that described damaged degree priority is maximum, Ω is the affected area in described image to be repaired, refer to the formwork that described damaged degree priority is maximum with the common factor of affected area Ω in described image to be repaired, the affected area in the formwork that namely described damaged degree priority is maximum; refer to that in the formwork that described damaged degree priority is maximum, get any pixel in affected area is for the pixel in damaged area in the formwork that described damaged degree priority is maximum, for pixel in damaged area in the formwork that described damaged degree priority is maximum degree of confidence.

The invention has the beneficial effects as follows: by introducing tensor theories in priority level initializing process, due to the effective Description Image partial structurtes information of structure tensor energy, relation between its eigenwert can the different subregion of token image, make it possible to according to its structure tensor and eigenwert thereof, control image repair priority, thus avoid in prior art because degree of confidence item is larger, and data item less time, the image block causing Given information a lot of can not get preferential reparation, produces mis repair.

On the basis of technique scheme, the present invention can also do following improvement:

Further, described step 5 comprises the following steps:

Step 5.1, travel through the sample block of all non-affected area in described image to be repaired, the size of the formwork that described sample block is maximum with described damaged degree priority is identical, and in judging the formwork that described damaged degree priority is maximum in all pixel value sums and described sample block all pixel value sums whether meet following relational expression:

(1-δ)·sum(Ψ _p”)≤sum(Ψ _q')≤(1+δ)·sum(Ψ _p”)；

In formula, p " be the known pixel of pixel in the maximum formwork of described damaged degree priority, Ψ _{p "}for the region that pixel in the formwork that described damaged degree priority is maximum is known; Sum (Ψ _{p "}) refer to that in the formwork maximum to described damaged degree priority, all pixel point values of pixel known portions are sued for peace; Q' is the corresponding pixel of pixel that in formwork maximum with described damaged degree priority in described sample block, pixel is known, Ψ _q'for the region that the region that pixel in formwork maximum with described damaged degree priority in described sample block is known is corresponding; Sum (Ψ _q') refer to that the pixel point value corresponding to known pixels point in formwork maximum with described damaged degree priority in described sample block is sued for peace; δ value is [0,1];

Step 5.2, after all pixel value sums meet the relational expression in step 5.1 in pixel value sums all in described sample block and the maximum formwork of the maximum described damaged degree priority of described damaged degree priority, carries out color-match;

Step 5.3, using sample block the highest for color-match degree as blocks and optimal matching blocks.

The beneficial effect of above-mentioned further scheme is adopted to be added to judge to be matched piece of known region pixel and area pixel value sum Satisfying Matching Conditions this process corresponding with object block before color-match, avoid in prior art and search for blocks and optimal matching blocks in global scope, increase time complexity, cause remediation efficiency this problem low.

Further, color-match is carried out according to following formula in described step 5.2:

${\mathrm{\Ψ}}_{\tilde{q}}=\arg \underset{{\mathrm{\Ψ}}_q\⋐\mathrm{\Φ}}{\min }d({\mathrm{\Ψ}}_{p_m},{\mathrm{\Ψ}}_q)+\underset{i=1}{\overset{n}{\Σ}}\[{({\mathrm{\λ}}_1-{\mathrm{\λ}}_1^{\′})}^2+{({\mathrm{\λ}}_2-{\mathrm{\λ}}_2^{\′})}^2\]$

In formula, refer to ask for make get the ψ of minimum value _q; λ ₁for the First Eigenvalue of the maximum formwork structure tensor expression formula of described damaged degree priority; λ ₂for the Second Eigenvalue of the maximum formwork structure tensor expression formula of described damaged degree priority; λ ' ₁for the First Eigenvalue of sample block structure tensor expression formula; λ ' ₂the Second Eigenvalue of sample block structure tensor expression formula; Φ is non-affected area in described image to be repaired; for the formwork that described damaged degree priority is maximum; Q is the pixel of the formwork center that in sample block, corresponding described damaged degree priority is maximum; ψ _qfor the sample block put centered by pixel q; N is the quantity of all pixels in described formwork; for the sample block that color-match degree in non-affected area is the highest;

the formwork maximum for described damaged degree priority and the color sum of squares of deviations of sample block, expression formula is:

$d({\mathrm{\Ψ}}_{p_m},{\mathrm{\Ψ}}_q)=\mathrm{\Σ}\[{(I_R-I_R^{\′})}^2+{(I_G-I_G^{\′})}^2+{(I_B-I_B^{\′})}^2\]$

In formula, I _rfor the gray-scale value of known pixels point in R passage in formwork; I _r' be the gray-scale value of known pixels point in R passage in sample block; I _gfor the gray-scale value of known pixels point in G passage in formwork; I _g' be the gray-scale value of known pixels point in G passage in sample block; I _bfor the gray-scale value of known pixels point in channel B in formwork; I _b' be the gray-scale value of known pixels point in channel B in sample block.

The beneficial effect of above-mentioned further scheme is adopted to be by the eigenwert of structure tensor is introduced matching criterior, to avoid in prior art random selecting object matching block in the middle of multiple coupling object block, cause choosing improper, and affect the effect of later image reparation, this color-match criterion makes found object block similarity higher, can reduce erroneous matching rate.

Further, described weight a ₁, a ₂, a ₃can arrange according to picture structure, when image texture characteristic to be repaired enriches, a ₁account for the largest percentage; When T-shaped partial structurtes abundant information such as turnings in image to be repaired, a ₂account for the largest percentage; When linear structure is enriched, a ₃account for the largest percentage.

Adopt the beneficial effect of above-mentioned further scheme to be proportion by suitably adjusting weight for the image to be repaired of different characteristics, can repairing quality be improved.

Further, described when image texture characteristic to be repaired enriches, a ₁: a ₂: a ₃=3:1:1; In described image to be repaired during the T-shaped partial structurtes abundant information such as turning, a ₁: a ₂: a ₃=1:3:1; Described when linear structure is enriched, a ₁: a ₂: a ₃=1:1:3.

Adopt the beneficial effect of above-mentioned further scheme to be proportion by arranging weight for the image to be repaired of different characteristics, can repairing quality be improved.

Present invention also offers a kind of Images uniting repair system based on partial structurtes feature, comprising:

Load module, for inputting an image to be repaired, is designated as u (x, y), and x, y are respectively the coordinate of image to be repaired on x, y direction;

Initialization module, to the degree of confidence initialization of all pixels in described image to be repaired; Specific as follows:

$C\left(p^{\′}\right)=0,\∀p^{\′}\∈\mathrm{\Ω},C\left(p^{\′}\right)=1,\∀p^{\′}\∈\mathrm{\Φ}$

In formula, p' is the pixel in described image to be repaired; The degree of confidence that C (p') is pixel p'; Ω is the affected area in described image to be repaired, and Φ is the non-affected area in described image to be repaired; for getting arbitrary value; refer to that the affected area in described image to be repaired gets any pixel; refer to that the non-affected area in described image to be repaired gets any pixel;

Formwork chooses module, for choosing the formwork of each pixel described respectively in described image damaged boundary to be repaired centered by each pixel;

Computing module, for for the size according to the damaged degree priority of each pixel place formwork in image damaged boundary to be repaired described in following formulae discovery, and by damaged degree priority according to descending sort from big to small;

P(p)＝a ₁C(p)+a ₂D(p)+a ₃H(p)；

In formula, p is the pixel of described formwork center; The damaged degree priority that P (p) is pixel p; a ₁, a ₂, a ₃for weight, and a ₁, a ₂, a ₃and be 1;

H (p)=Kh+exp (-h); K is controling parameter, and value is 0.8; Exp (-h)=e ^-h, e ^-hfor taking e as the exponential function of the truth of a matter, e=2.71828; H is local metric function, h=| λ ₁-λ ₂| ², λ ₁and λ ₂be respectively the First Eigenvalue and the Second Eigenvalue of described formwork structure tensor expression formula, expression formula is:

${\mathrm{\λ}}_{1,2}=\frac{1}{2}(j_{11}+j_{22}\±\sqrt{{(j_{11}+j_{22})}^2+4j_{12}^2})$

Wherein, j ₁₁, j ₁₂, j ₂₂obtained by its structure tensor expression formula to be repaired, described its structure tensor expression formula to be repaired is:

ask convolution algorithm; refer to G _ρwith ask convolution;

refer to G _ρwith ask convolution algorithm;

C (p) is the degree of confidence of described formwork center pixel p; D (p) is the data item of described formwork center pixel p;

$C\left(p\right)=\frac{\underset{p^{\′}\∈{\mathrm{\Ψ}}_p\∩\mathrm{\Ω}}{\Σ}C\left(p^{\′}\right)}{\left|{\mathrm{\Ψ}}_p\right|}$

$D\left(p\right)=\frac{|\▿I_p^{\⊥}\·n_p|}{\mathrm{\α}}$

In formula, ψ _pfor the formwork that damaged boundary is put centered by pixel p, Φ is the non-affected area in described image to be repaired, ψ _pthe common factor of the non-affected area in ∩ Φ finger print plate and described image to be repaired, for the pixel in the common factor of the affected area in described formwork and described image to be repaired, i.e. the pixel of non-affected area in described formwork, for pixel degree of confidence, | Ψ _p| be the area of formwork, the quantity of pixel in finger print plate; for the direction of the isophote of pixel p, i.e. the vertical direction of gradient, n _pfor the unit direction vector of pixel p, α is normalized factor, and in gray level image, value is 255; $C\left(\hat{p}\right)=C\left(p^{\′}\right)=1,\∀p^{\′}\∈\mathrm{\Φ};$

Matching module, for finding the highest region of the formwork similarity maximum with described damaged degree priority in the non-affected area of described image to be repaired, as blocks and optimal matching blocks;

Replication module, for copying in the maximum formwork of described damaged degree priority by the pixel value in described blocks and optimal matching blocks;

Update module, for upgrade the maximum formwork of described damaged degree priority according to following formula damaged area in the degree of confidence of all pixels, and return computing module, until image to be repaired is repaired completely;

$C\left(\hat{q}\right)C\left(p_m\right),\∀\hat{q}\∈{\mathrm{\Ψ}}_{p_m}\∩\mathrm{\Ω};$

In formula, p _mfor the pixel of the center of the maximum formwork of described damaged degree priority; C (p _m) be pixel p _mdegree of confidence, for the formwork that described damaged degree priority is maximum, Ω is the affected area in described image to be repaired, refer to the formwork that described damaged degree priority is maximum with the common factor of affected area Ω in described image to be repaired, the affected area in the formwork that namely described damaged degree priority is maximum; refer to that in the formwork that described damaged degree priority is maximum, get any pixel in affected area is for the pixel in damaged area in the formwork that described damaged degree priority is maximum, for pixel in damaged area in the formwork that described damaged degree priority is maximum degree of confidence.

The invention has the beneficial effects as follows: by introducing tensor theories in priority level initializing process, due to the effective Description Image partial structurtes information of structure tensor energy, relation between its eigenwert can the different subregion of token image, make it possible to according to its structure tensor and eigenwert thereof, control image repair priority, thus avoid in prior art because degree of confidence item is larger, and data item less time, the image block causing Given information a lot of can not get preferential reparation, produces mis repair.

On the basis of technique scheme, the present invention can also do following improvement:

Further, described matching module comprises:

Judging unit, for traveling through the sample block of all non-affected area in described image to be repaired, the size of the formwork that described sample block is maximum with described damaged degree priority is identical, and in judging the formwork that described damaged degree priority is maximum in all pixel value sums and described sample block all pixel value sums whether meet following relational expression:

(1-δ)·sum(Ψ _p”)≤sum(Ψ _q')≤(1+δ)·sum(Ψ _p”)；

In formula, p " be the known pixel of pixel in the maximum formwork of described damaged degree priority, Ψ _{p "}for the region that pixel in the formwork that described damaged degree priority is maximum is known; Sum (Ψ _{p "}) refer to that in the formwork maximum to described damaged degree priority, all pixel point values of pixel known portions are sued for peace; Q' is the corresponding pixel of pixel that in formwork maximum with described damaged degree priority in described sample block, pixel is known, Ψ _q'for the region that the region that pixel in formwork maximum with described damaged degree priority in described sample block is known is corresponding; Sum (Ψ _q') refer to that the pixel point value corresponding to known pixels point in formwork maximum with described damaged degree priority in described sample block is sued for peace; δ value is [0,1];

Color-match unit, for after all pixel value sums meet the relational expression in judging unit in pixel value sums all in described sample block and the maximum formwork of the maximum described damaged degree priority of described damaged degree priority, carries out color-match;

Optimum Matching unit, for using sample block the highest for color-match degree as blocks and optimal matching blocks.

The beneficial effect of above-mentioned further scheme is adopted to be added to judge to be matched piece of known region pixel and area pixel value sum Satisfying Matching Conditions this process corresponding with object block before color-match, avoid in prior art and search for blocks and optimal matching blocks in global scope, increase time complexity, cause remediation efficiency this problem low.

Further, described color-match unit carries out color-match according to following formula:

${\mathrm{\Ψ}}_{\tilde{q}}=\arg \underset{{\mathrm{\Ψ}}_q\⋐\mathrm{\Φ}}{\min }d({\mathrm{\Ψ}}_{p_m},{\mathrm{\Ψ}}_q)+\underset{i=1}{\overset{n}{\Σ}}\[{({\mathrm{\λ}}_1-{\mathrm{\λ}}_1^{\′})}^2+{({\mathrm{\λ}}_2-{\mathrm{\λ}}_2^{\′})}^2\]$

In formula, refer to ask for make get the ψ of minimum value _q; λ ₁for the First Eigenvalue of the maximum formwork structure tensor expression formula of described damaged degree priority; λ ₂for the Second Eigenvalue of the maximum formwork structure tensor expression formula of described damaged degree priority; λ ' ₁for the First Eigenvalue of sample block structure tensor expression formula; λ ' ₂the Second Eigenvalue of sample block structure tensor expression formula; Φ is non-affected area in described image to be repaired; for the formwork that described damaged degree priority is maximum; Q is the pixel of the formwork center that in sample block, corresponding described damaged degree priority is maximum; ψ _qfor the sample block put centered by pixel q; N is the quantity of all pixels in described formwork; for the sample block that color-match degree in non-affected area is the highest;

the formwork maximum for described damaged degree priority and the color sum of squares of deviations of sample block, expression formula is:

$d({\mathrm{\Ψ}}_{p_m},{\mathrm{\Ψ}}_q)=\mathrm{\Σ}\[{(I_R-I_R^{\′})}^2+{(I_G-I_G^{\′})}^2+{(I_B-I_B^{\′})}^2\]$

In formula, I _rfor the gray-scale value of known pixels point in R passage in formwork; I _r' be the gray-scale value of known pixels point in R passage in sample block; I _gfor the gray-scale value of known pixels point in G passage in formwork; I _g' be the gray-scale value of known pixels point in G passage in sample block; I _bfor the gray-scale value of known pixels point in channel B in formwork; I _b' be the gray-scale value of known pixels point in channel B in sample block.

The beneficial effect of above-mentioned further scheme is adopted to be by the eigenwert of structure tensor is introduced matching criterior, to avoid in prior art random selecting object matching block in the middle of multiple coupling object block, cause choosing improper, and affect the effect of later image reparation, this color-match criterion makes found object block similarity higher, can reduce erroneous matching rate.

Further, described computing module comprises setting unit, for arranging weight a according to picture structure ₁, a ₂, a ₃, when image texture characteristic to be repaired enriches, a ₁account for the largest percentage; When T-shaped partial structurtes abundant information such as turnings in image to be repaired, a ₂account for the largest percentage; When linear structure is enriched, a ₃account for the largest percentage.

Adopt the beneficial effect of above-mentioned further scheme to be proportion by suitably adjusting weight for the image to be repaired of different characteristics, can repairing quality be improved.

Further, described setting unit comprises and arranges subelement, for when image texture characteristic to be repaired enriches, and a ₁: a ₂: a ₃=3:1:1; In described image to be repaired during the T-shaped partial structurtes abundant information of corner, a ₁: a ₂: a ₃=1:3:1; Described when linear structure is enriched, a ₁: a ₂: a ₃=1:1:3.

Adopt the beneficial effect of above-mentioned further scheme to be proportion by arranging weight for the image to be repaired of different characteristics, can repairing quality be improved.

Accompanying drawing explanation

Fig. 1 is the schematic flow sheet of a kind of Images uniting restorative procedure based on partial structurtes feature of the present invention;

Fig. 2 is the Comparative result figure utilizing existing method and the inventive method to repair first group of image during the present invention tests;

Fig. 3 is the Comparative result figure utilizing existing method and the inventive method to repair second group of image during the present invention tests;

Fig. 4 is the Comparative result figure utilizing existing method and the inventive method to repair the 3rd group of image during the present invention tests;

Fig. 5 is the Comparative result figure utilizing existing method and the inventive method to repair the 4th group of image during the present invention tests;

Fig. 6 is the Comparative result figure utilizing existing method and the inventive method to repair the 5th group of image during the present invention tests.

Embodiment

Be described principle of the present invention and feature below in conjunction with accompanying drawing, example, only for explaining the present invention, is not intended to limit scope of the present invention.

As shown in Figure 1, a kind of Images uniting restorative procedure based on partial structurtes feature of the present invention, comprises the following steps:

Step 1, input an image to be repaired, be designated as u (x, y), x, y are respectively the coordinate of image to be repaired on x, y direction;

To the degree of confidence initialization of all pixels in described image to be repaired; Specific as follows:

$C\left(p^{\′}\right)=0,\∀p^{\′}\∈\mathrm{\Ω},C\left(p^{\′}\right)=1,\∀p^{\′}\∈\mathrm{\Φ}$

In formula, p' is the pixel in described image to be repaired; The degree of confidence that C (p') is pixel p'; Ω is the affected area in described image to be repaired, and Φ is the non-affected area in described image to be repaired; for getting arbitrary value; refer to that the affected area in described image to be repaired gets any pixel; refer to that the non-affected area in described image to be repaired gets any pixel;

Step 3, chooses the formwork of each pixel described respectively in described image damaged boundary to be repaired centered by each pixel;

Step 4, according to the size of the damaged degree priority of each pixel place formwork in image damaged boundary to be repaired described in following formulae discovery, and by damaged degree priority according to descending sort from big to small;

P(p)＝a ₁C(p)+a ₂D(p)+a ₃H(p)；

Wherein, p is the pixel of described formwork center; The damaged degree priority that P (p) is pixel p; a ₁, a ₂, a ₃for weight, and a ₁, a ₂, a ₃and be 1;

H (p)=Kh+exp (-h); K is controling parameter, and value is 0.8; Exp (-h)=e ^-h, e ^-hfor taking e as the exponential function of the truth of a matter, e=2.71828; H is local metric function, h=| λ ₁-λ ₂| ², λ ₁and λ ₂be respectively the First Eigenvalue and the Second Eigenvalue of described formwork structure tensor expression formula, expression formula is:

${\mathrm{\λ}}_{1,2}=\frac{1}{2}(j_{11}+j_{22}\±\sqrt{{(j_{11}+j_{22})}^2+4j_{12}^2})$

Wherein, j ₁₁, j ₁₂, j ₂₂obtained by its structure tensor expression formula to be repaired, described its structure tensor expression formula to be repaired is:

refer to G _ρwith ask convolution algorithm; refer to G _ρwith ask convolution algorithm;

refer to G _ρwith ask convolution algorithm;

C (p) is the degree of confidence of described formwork center pixel p; D (p) is the data item of described formwork center pixel p;

$C\left(p\right)=\frac{\underset{\hat{p}\∈{\mathrm{\Ψ}}_p\∩\mathrm{\Φ}}{\Σ}C\left(\hat{p}\right)}{\left|{\mathrm{\Ψ}}_p\right|}$

$D\left(p\right)=\frac{|\▿I_p^{\⊥}\·n_p|}{\mathrm{\α}}$

In formula, ψ _pfor the formwork that damaged boundary is put centered by pixel p, Φ is the non-affected area in described image to be repaired, ψ _pthe common factor of the non-affected area in ∩ Φ finger print plate and described image to be repaired, for the pixel in the common factor of the non-affected area in described formwork and described image to be repaired, i.e. the pixel of non-affected area in described formwork, for pixel degree of confidence, | Ψ _p| be the area of formwork, the quantity of pixel in finger print plate; for the direction of the isophote of pixel p, i.e. the vertical direction of gradient, n _pfor the unit direction vector of pixel p, α is normalized factor, and in gray level image, value is 255; $C\left(\hat{p}\right)=C\left(p^{\′}\right)=1,\∀p^{\′}\∈\mathrm{\Φ};$

In H (p), add exp (-h), have following two advantages: when two eigenwert differences are little, when namely image is tending towards flat site, ensure the very little and non-vanishing energy again of H (p) change; When difference between two eigenwerts is increasing, its effect is but more and more less, ensures the linear followability of H (p), with the change of accurate description picture structure;

H (p) increases the impact of structural factor in repair process on the one hand, ensure that the preferential propagation of picture structure; Introduce on the other hand structure tensor make the description of picture structure and judge more accurate, thus ensure that repairing effect.

Step 5, finds the formwork maximum with described damaged degree priority in the non-affected area of described image to be repaired the region that similarity is the highest, as blocks and optimal matching blocks;

Step 5 comprises the following steps:

Step 5.1, travel through the sample block of all non-affected area in described image to be repaired, the size of the formwork that described sample block is maximum with described damaged degree priority is identical, and in judging the formwork that described damaged degree priority is maximum in all pixel value sums and described sample block all pixel value sums whether meet following relational expression:

(1-δ)·sum(Ψ _p”)≤sum(Ψ _q')≤(1+δ)·sum(Ψ _p”)；

In formula, p " be the known pixel of pixel in the maximum formwork of described damaged degree priority, Ψ _{p "}for the region that pixel in the formwork that described damaged degree priority is maximum is known; Sum (Ψ _{p "}) refer to that in the formwork maximum to described damaged degree priority, all pixel point values of pixel known portions are sued for peace; Q' is the corresponding pixel of pixel that in formwork maximum with described damaged degree priority in described sample block, pixel is known, Ψ _q'for the region that the region that pixel in formwork maximum with described damaged degree priority in described sample block is known is corresponding; Sum (Ψ _q') refer to that the pixel point value corresponding to known pixels point in formwork maximum with described damaged degree priority in described sample block is sued for peace; δ value is [0,1];

Step 5.2, after all pixel value sums meet the relational expression in step 5.1 in pixel value sums all in described sample block and the maximum formwork of the maximum described damaged degree priority of described damaged degree priority, carries out color-match;

Color-match is carried out according to following formula in step 5.2:

${\mathrm{\Ψ}}_{\tilde{q}}=\arg \underset{{\mathrm{\Ψ}}_q\⋐\mathrm{\Φ}}{\min }d({\mathrm{\Ψ}}_{p_m},{\mathrm{\Ψ}}_q)+\underset{i=1}{\overset{n}{\Σ}}\[{({\mathrm{\λ}}_1-{\mathrm{\λ}}_1^{\′})}^2+{({\mathrm{\λ}}_2-{\mathrm{\λ}}_2^{\′})}^2\]$

In formula, refer to ask for make get the ψ of minimum value _q; λ ₁for the First Eigenvalue of the maximum formwork structure tensor expression formula of described damaged degree priority; λ ₂for the Second Eigenvalue of the maximum formwork structure tensor expression formula of described damaged degree priority; λ ' ₁for the First Eigenvalue of sample block structure tensor expression formula; λ ' ₂the Second Eigenvalue of sample block structure tensor expression formula; Φ is non-affected area in described image to be repaired; for the formwork that described damaged degree priority is maximum; Q is the pixel of the formwork center that in sample block, corresponding described damaged degree priority is maximum; ψ _qfor the sample block put centered by pixel q; N is the quantity of all pixels in described formwork; for the sample block that color-match degree in non-affected area is the highest;

the formwork maximum for described damaged degree priority and the color sum of squares of deviations of sample block, expression formula is:

$d({\mathrm{\Ψ}}_{p_m},{\mathrm{\Ψ}}_q)=\mathrm{\Σ}\[{(I_R-I_R^{\′})}^2+{(I_G-I_G^{\′})}^2+{(I_B-I_B^{\′})}^2\]$

In formula, I _rfor the gray-scale value of known pixels point in R passage in formwork; I _r' be the gray-scale value of known pixels point in R passage in sample block; I _gfor the gray-scale value of known pixels point in G passage in formwork; I _g' be the gray-scale value of known pixels point in G passage in sample block; I _bfor the gray-scale value of known pixels point in channel B in formwork; I _b' be the gray-scale value of known pixels point in channel B in sample block.

Step 5.3, using sample block the highest for color-match degree as blocks and optimal matching blocks;

Step 6, copies to the formwork that described damaged degree priority is maximum by the pixel value in described blocks and optimal matching blocks in;

Step 7, upgrades the maximum formwork of described damaged degree priority according to following formula damaged area in the degree of confidence of all pixels, and return step 4, until image to be repaired is repaired completely;

$C\left(\hat{q}\right)=C\left(p_m\right),\∀\hat{p}\∈{\mathrm{\Ψ}}_{p_m}\∩\mathrm{\Ω};$

In formula, p _mfor the pixel of the center of the maximum formwork of described damaged degree priority; C (p _m) be pixel p _mdegree of confidence, for the formwork that described damaged degree priority is maximum, Ω is the affected area in described image to be repaired, refer to the formwork that described damaged degree priority is maximum with the common factor of affected area Ω in described image to be repaired, the affected area in the formwork that namely described damaged degree priority is maximum; refer to that in the formwork that described damaged degree priority is maximum, get any pixel in affected area is for the pixel in damaged area in the formwork that described damaged degree priority is maximum, for pixel in damaged area in the formwork that described damaged degree priority is maximum degree of confidence.

Present invention also offers a kind of Images uniting repair system based on partial structurtes feature, comprising:

Load module, for inputting an image to be repaired, is designated as u (x, y), and x, y are respectively the coordinate of image to be repaired on x, y direction;

Initialization module, for the degree of confidence initialization to all pixels in described image to be repaired; Specific as follows:

$C\left(p^{\′}\right)=0,\∀p^{\′}\∈\mathrm{\Ω},C\left(p^{\′}\right)=1,\∀p^{\′}\∈\mathrm{\Φ}$

In formula, p' is the pixel in described image to be repaired; The degree of confidence that C (p') is pixel p'; Ω is the affected area in described image to be repaired, and Φ is the non-affected area in described image to be repaired; for getting arbitrary value; refer to that the affected area in described image to be repaired gets any pixel; refer to that the non-affected area in described image to be repaired gets any pixel;

Formwork chooses module, for choosing the formwork of each pixel described respectively in described image damaged boundary to be repaired centered by each pixel;

Computing module, for the size according to the damaged degree priority of each pixel place formwork in image damaged boundary to be repaired described in following formulae discovery, and by damaged degree priority according to descending sort from big to small;

P(p)＝a ₁C(p)+a ₂D(p)+a ₃H(p)；

In formula, p is the pixel of described formwork center; The damaged degree priority that P (p) is pixel p; a ₁, a ₂, a ₃for weight, and and be 1;

H (p)=Kh+exp (-h); K is controling parameter, and value is 0.8; Exp (-h)=e ^-h, e ^-hfor taking e as the exponential function of the truth of a matter, e=2.71828; H is local metric function, h=| λ ₁-λ ₂| ², λ ₁and λ ₂be respectively the First Eigenvalue and the Second Eigenvalue of described formwork structure tensor expression formula, expression formula is:

${\mathrm{\λ}}_{1,2}=\frac{1}{2}(j_{11}+j_{22}\±\sqrt{{(j_{11}+j_{22})}^2+4j_{12}^2})$

Wherein, j ₁₁, j ₁₂, j ₂₂obtained by its structure tensor expression formula to be repaired, described its structure tensor expression formula to be repaired is:

refer to G _ρwith ask convolution algorithm; refer to G _ρwith ask convolution algorithm;

refer to G _ρwith ask convolution algorithm;

C (p) is the degree of confidence of described formwork center pixel p; D (p) is the data item of described formwork center pixel p;

$C\left(p\right)=\frac{\underset{p^{\′}\∈{\mathrm{\Ψ}}_p\∩\mathrm{\Ω}}{\Σ}C\left(p^{\′}\right)}{\left|{\mathrm{\Ψ}}_p\right|}$

$D\left(p\right)=\frac{|\▿I_p^{\⊥}\·n_p|}{\mathrm{\α}}$

In formula, ψ _pfor the formwork that damaged boundary is put centered by pixel p, Φ is the non-affected area in described image to be repaired, ψ _pthe common factor of the non-affected area in ∩ Φ finger print plate and described image to be repaired, for the pixel in the common factor of the affected area in described formwork and described image to be repaired, i.e. the pixel of non-affected area in described formwork, for pixel degree of confidence, | Ψ _p| be the area of formwork, the quantity of pixel in finger print plate; for the direction of the isophote of pixel p, i.e. the vertical direction of gradient, n _pfor the unit direction vector of pixel p, α is normalized factor, and in gray level image, value is 255; $C\left(\hat{p}\right)=C\left(p^{\′}\right)=1,\∀p^{\′}\∈\mathrm{\Φ};$

Wherein, described computing module comprises setting unit, for arranging weight a according to picture structure ₁, a ₂, a ₃, when image texture characteristic to be repaired enriches, a ₁account for the largest percentage; When T-shaped partial structurtes abundant information such as turnings in image to be repaired, a ₂account for the largest percentage; When linear structure is enriched, a ₃account for the largest percentage;

Described setting unit also for when image texture characteristic to be repaired enriches, a ₁: a ₂: a ₃=3:1:1; In described image to be repaired during the T-shaped partial structurtes abundant information of corner, a ₁: a ₂: a ₃=1:3:1; Described when linear structure is enriched, a ₁: a ₂: a ₃=1:1:3.

Matching module, for finding the highest region of the formwork similarity maximum with described damaged degree priority in the non-affected area of described image to be repaired, as blocks and optimal matching blocks.

Described matching module comprises:

Judging unit, for traveling through the sample block of all non-affected area in described image to be repaired, the size of the formwork that described sample block is maximum with described damaged degree priority is identical, and in judging the formwork that described damaged degree priority is maximum in all pixel value sums and described sample block all pixel value sums whether meet following relational expression:

(1-δ)·sum(Ψ _p”)≤sum(Ψ _q')≤(1+δ)·sum(Ψ _p”)；

In formula, p " be the known pixel of pixel in the maximum formwork of described damaged degree priority, Ψ _{p "}for the region that pixel in the formwork that described damaged degree priority is maximum is known; Sum (Ψ _{p "}) refer to that in the formwork maximum to described damaged degree priority, all pixel point values of pixel known portions are sued for peace; Q' is the corresponding pixel of pixel that in formwork maximum with described damaged degree priority in described sample block, pixel is known, Ψ _q'for the region that the region that pixel in formwork maximum with described damaged degree priority in described sample block is known is corresponding; Sum (Ψ _q') refer to that the pixel point value corresponding to known pixels point in formwork maximum with described damaged degree priority in described sample block is sued for peace; δ value is [0,1];

Color-match unit, for after all pixel value sums meet the relational expression in judging unit in pixel value sums all in described sample block and the maximum formwork of the maximum described damaged degree priority of described damaged degree priority, carries out color-match;

Described color-match unit carries out color-match according to following formula:

${\mathrm{\Ψ}}_{\tilde{q}}=\arg \underset{{\mathrm{\Ψ}}_q\⋐\mathrm{\Φ}}{\min }d({\mathrm{\Ψ}}_{p_m},{\mathrm{\Ψ}}_q)+\underset{i=1}{\overset{n}{\Σ}}\[{({\mathrm{\λ}}_1-{\mathrm{\λ}}_1^{\′})}^2+{({\mathrm{\λ}}_2-{\mathrm{\λ}}_2^{\′})}^2\]$

In formula, refer to ask for make get the ψ of minimum value _q; λ ₁for the First Eigenvalue of the maximum formwork structure tensor expression formula of described damaged degree priority; λ ₂for the Second Eigenvalue of the maximum formwork structure tensor expression formula of described damaged degree priority; λ ' ₁for the First Eigenvalue of sample block structure tensor expression formula; λ ' ₂the Second Eigenvalue of sample block structure tensor expression formula; Φ is non-affected area in described image to be repaired; for the formwork that described damaged degree priority is maximum; Q is the pixel of the formwork center that in sample block, corresponding described damaged degree priority is maximum; ψ _qfor the sample block put centered by pixel q; N is the quantity of all pixels in described formwork; for the sample block that color-match degree in non-affected area is the highest;

the formwork maximum for described damaged degree priority and the color sum of squares of deviations of sample block, expression formula is:

$d({\mathrm{\Ψ}}_{p_m},{\mathrm{\Ψ}}_q)=\mathrm{\Σ}\[{(I_R-I_R^{\′})}^2+{(I_G-I_G^{\′})}^2+{(I_B-I_B^{\′})}^2\]$

In formula, I _rfor the gray-scale value of known pixels point in R passage in formwork; I _r' be the gray-scale value of known pixels point in R passage in sample block; I _gfor the gray-scale value of known pixels point in G passage in formwork; I _g' be the gray-scale value of known pixels point in G passage in sample block; I _bfor the gray-scale value of known pixels point in channel B in formwork; I _b' be the gray-scale value of known pixels point in channel B in sample block.

Replication module, for copying to the pixel value in described blocks and optimal matching blocks in the maximum formwork of described damaged degree priority.

Update module, for upgrade the maximum formwork of described damaged degree priority according to following formula damaged area in the degree of confidence of all pixels, and return computing module, until image to be repaired is repaired completely;

$C\left(\hat{q}\right)=C\left(p_m\right),\∀\hat{q}\∈{\mathrm{\Ψ}}_{p_m}\∩\mathrm{\Ω};$

In formula, p _mfor the pixel of the center of the maximum formwork of described damaged degree priority; C (p _m) be pixel p _mdegree of confidence, for the formwork that described damaged degree priority is maximum, Ω is the affected area in described image to be repaired, refer to the formwork that described damaged degree priority is maximum with the common factor of affected area Ω in described image to be repaired, the affected area in the formwork that namely described damaged degree priority is maximum; refer to that in the formwork that described damaged degree priority is maximum, get any pixel in affected area is for the pixel in damaged area in the formwork that described damaged degree priority is maximum, for pixel in damaged area in the formwork that described damaged degree priority is maximum degree of confidence.

Optimum Matching unit, for using sample block the highest for color-match degree as blocks and optimal matching blocks.

Below by experiment, effect of the present invention is described:

Fig. 2, Fig. 3, Fig. 4, Fig. 5 and Fig. 6 is respectively the Comparative result figure utilizing existing method and the inventive method to repair image to be repaired in the present invention's experiment, figure (a1), figure (b1), figure (c1), figure (d1) and figure (e1) is original image, figure (a2), figure (b2), figure (c2), figure (d2) and figure (e2) is image to be repaired, figure (a3), figure (b3), figure (c3), figure (d3) and figure (e3) is respectively and utilizes Criminsi algorithm of the prior art to carry out the result figure repaired, figure (a4), figure (b4), figure (c4), figure (d4) and figure (e4) are respectively and utilize CDD model restore design to carry out the result figure repaired, figure (a5), figure (b5), figure (c5), figure (d5) and figure (e5) are respectively and utilize the image repair algorithm based on textures synthesis improved to carry out the result figure repaired, figure (a6), figure (b6), figure (c6), figure (d6) and figure (e6) are respectively and utilize method of the present invention to carry out the result figure repaired.

Table 4.1, table 4.2, table 4.3 and table 4.4 are respectively and utilize existing method and method of the present invention to start with from evaluation criterion to contrast, and evaluation criterion comprises: Y-PSNR (PSNR), absolute mean error (MAE), structural similarity (MSSIM) and program runtime (s); Shown in specific as follows:

PSNR (dB) value of all kinds of algorithm of table 4.1 compares

The MAE value of all kinds of algorithm of table 4.2 compares

The MSSIM value of all kinds of algorithm of table 4.3 compares

The working time (s) of all kinds of algorithm of table 4.4 compares

Every experimental data shows that the inventive method is in all kinds of objective evaluation index above, i.e. PSNR value, MAE value structural similarity (SSIM) and be all better than other several existing methods working time.Mainly due to following 3 points: the 1) introducing of picture structure measure function, and priority becomes the pro forma transformation of addition from being multiplied, ensure that and the propagation that image structure information is correct and preferential improve repairing effect; 2) introducing of structure tensor eigenwert makes mating of multiblock to be repaired and object block more accurate, reduces erroneous matching rate and the cumulative rate of mistake of former algorithm, visual effect is had obvious improvement; 3) transformation of way of search, decreases the redundance of to be matched piece, reduces time complexity, and remediation efficiency is significantly increased.

These are only preferred embodiment of the present invention, not in order to limit the present invention, within the spirit and principles in the present invention all, any amendment done, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (10)

1., based on an Images uniting restorative procedure for partial structurtes feature, it is characterized in that, comprise the following steps:

Step 1, input an image to be repaired, be designated as u (x, y), x, y are respectively the coordinate of image to be repaired on x, y direction;

Step 2, to the degree of confidence initialization of all pixels in described image to be repaired; Specific as follows:

$C\left(p^{\′}\right)=0,\∀p^{\′}\∈\mathrm{\Ω},C\left(p^{\′}\right)=1,\∀p^{\′}\∈\mathrm{\Φ}$

In formula, p' is the pixel in described image to be repaired; The degree of confidence that C (p') is pixel p'; Ω is the affected area in described image to be repaired, and Φ is the non-affected area in described image to be repaired; for getting arbitrary value; refer to that the affected area in described image to be repaired gets any pixel; refer to that the non-affected area in described image to be repaired gets any pixel;

Step 3, chooses the formwork of each pixel described respectively in described image damaged boundary to be repaired centered by each pixel;

Step 4, according to the size of the damaged degree priority of each pixel place formwork in image damaged boundary to be repaired described in following formulae discovery, and by damaged degree priority according to descending sort from big to small;

P(p)＝a ₁C(p)+a ₂D(p)+a ₃H(p)；

Wherein, p is the pixel of described formwork center; The damaged degree priority that P (p) is pixel p; a ₁, a ₂, a ₃for weight, and a ₁, a ₂, a ₃and be 1;

H (p)=Kh+exp (-h); K is controling parameter, and value is 0.8; Exp (-h)=e ^-h, e ^-hfor taking e as the exponential function of the truth of a matter, e=2.71828; H is local metric function, h=| λ ₁-λ ₂| ², λ ₁and λ ₂be respectively the First Eigenvalue and the Second Eigenvalue of described formwork structure tensor expression formula, expression formula is:

${\mathrm{\λ}}_{1,2}=\frac{1}{2}(j_{11}+j_{22}\±\sqrt{{(j_{11}+j_{22})}^2+4j_{12}^2})$

Wherein, j ₁₁, j ₁₂, j ₂₂obtained by its structure tensor expression formula to be repaired, described its structure tensor expression formula to be repaired is:

Refer to G _ρwith ask convolution algorithm; refer to G _ρwith ask convolution algorithm; refer to G _ρwith ask convolution algorithm;

C (p) is the degree of confidence of described formwork center pixel p; D (p) is the data item of described formwork center pixel p;

$C\left(p\right)=\frac{\underset{\hat{p}\∈{\mathrm{\Ψ}}_P\∩\mathrm{\Φ}}{\Σ}C\left(\hat{p}\right)}{\left|{\mathrm{\Ψ}}_p\right|}$

$D\left(p\right)=\frac{|\▿I_p^{\⊥}\·n_p|}{\mathrm{\α}}$

In formula, ψ _pfor the formwork that damaged boundary is put centered by pixel p, Φ is the non-affected area in described image to be repaired, ψ _pthe common factor of the non-affected area in ∩ Φ finger print plate and described image to be repaired, for the pixel in the common factor of the non-affected area in described formwork and described image to be repaired, i.e. the pixel of non-affected area in described formwork, for pixel degree of confidence, | Ψ _p| be the area of formwork, the quantity of pixel in finger print plate; for the direction of the isophote of pixel p, i.e. the vertical direction of gradient, n _pfor the unit direction vector of pixel p, α is normalized factor, and in gray level image, value is 255; $C\left(\hat{p}\right)=C\left(p^{\′}\right)=1,\∀p^{\′}\∈\mathrm{\Φ};$

Step 5, finds the formwork maximum with described damaged degree priority in the non-affected area of described image to be repaired the region that similarity is the highest, as blocks and optimal matching blocks;

Step 6, copies to the formwork that described damaged degree priority is maximum by the pixel value in described blocks and optimal matching blocks in;

Step 7, upgrades the maximum formwork of described damaged degree priority according to following formula damaged area in the degree of confidence of all pixels, and return step 4, until image to be repaired is repaired completely;

$\begin{array}{cc}C\left(\hat{q}\right)=C\left(p_m\right)& \∀\hat{q}\∈{\mathrm{\Ψ}}_{p_m}\∩\mathrm{\Ω};\end{array}$

In formula, p _mfor the pixel of the center of the maximum formwork of described damaged degree priority; C (p _m) be pixel p _mdegree of confidence, for the formwork that described damaged degree priority is maximum, Ω is the affected area in described image to be repaired, refer to the formwork that described damaged degree priority is maximum with the common factor of affected area Ω in described image to be repaired, the affected area in the formwork that namely described damaged degree priority is maximum; refer to that in the formwork that described damaged degree priority is maximum, get any pixel in affected area is for the pixel in damaged area in the formwork that described damaged degree priority is maximum, for pixel in damaged area in the formwork that described damaged degree priority is maximum degree of confidence.

2. a kind of Images uniting restorative procedure based on partial structurtes feature according to claim 1, it is characterized in that, described step 5 comprises the following steps:

Step 5.1, travel through the sample block of all non-affected area in described image to be repaired, the size of the formwork that described sample block is maximum with described damaged degree priority is identical, and in judging the formwork that described damaged degree priority is maximum in all pixel value sums and described sample block all pixel value sums whether meet following relational expression:

(1-δ)·sum(Ψ _p”)≤sum(Ψ _q')≤(1+δ)·sum(Ψ _p”)；

In formula, p " be the known pixel of pixel in the maximum formwork of described damaged degree priority, Ψ _{p "}for the region that pixel in the formwork that described damaged degree priority is maximum is known; Sum (Ψ _{p "}) refer to that in the formwork maximum to described damaged degree priority, all pixel point values of pixel known portions are sued for peace; Q' is the corresponding pixel of pixel that in formwork maximum with described damaged degree priority in described sample block, pixel is known, Ψ _q'for the region that the region that pixel in formwork maximum with described damaged degree priority in described sample block is known is corresponding; Sum (Ψ _q') refer to that the pixel point value corresponding to known pixels point in formwork maximum with described damaged degree priority in described sample block is sued for peace; δ value is [0,1];

Step 5.2, after all pixel value sums meet the relational expression in step 5.1 in pixel value sums all in described sample block and the maximum formwork of the maximum described damaged degree priority of described damaged degree priority, carries out color-match;

Step 5.3, using sample block the highest for color-match degree as blocks and optimal matching blocks.

3., according to a kind of Images uniting restorative procedure based on partial structurtes feature described in claim 2, it is characterized in that, in described step 5.2, carry out color-match according to following formula:

${\mathrm{\Ψ}}_{\tilde{q}}=\arg \underset{{\mathrm{\Ψ}}_q\⋐\mathrm{\Φ}}{\min }d({\mathrm{\Ψ}}_{p_m},{\mathrm{\Ψ}}_q)+\underset{i=1}{\overset{n}{\Σ}}\[{({\mathrm{\λ}}_1-{\mathrm{\λ}}_1^{\′})}^2+{({\mathrm{\λ}}_2-{\mathrm{\λ}}_2^{\′})}^2\]$

In formula, refer to ask for make get the ψ of minimum value _q; λ ₁for the First Eigenvalue of the maximum formwork structure tensor expression formula of described damaged degree priority; λ ₂for the Second Eigenvalue of the maximum formwork structure tensor expression formula of described damaged degree priority; λ ' ₁for the First Eigenvalue of sample block structure tensor expression formula; λ ' ₂the Second Eigenvalue of sample block structure tensor expression formula; Φ is non-affected area in described image to be repaired; for the formwork that described damaged degree priority is maximum; Q is the pixel of the formwork center that in sample block, corresponding described damaged degree priority is maximum; ψ _qfor the sample block put centered by pixel q; N is the quantity of all pixels in described formwork; for the sample block that color-match degree in non-affected area is the highest;

the formwork maximum for described damaged degree priority and the color sum of squares of deviations of sample block, expression formula is:

$d({\mathrm{\Ψ}}_{p_m},{\mathrm{\Ψ}}_q)=\mathrm{\Σ}\[{(I_R-I_R^{\′})}^2+{(I_G-I_G^{\′})}^2+{(I_B-I_B^{\′})}^2\]$

In formula, I _rfor the gray-scale value of known pixels point in R passage in formwork; I _r' be the gray-scale value of known pixels point in R passage in sample block; I _gfor the gray-scale value of known pixels point in G passage in formwork; I _g' be the gray-scale value of known pixels point in G passage in sample block; I _bfor the gray-scale value of known pixels point in channel B in formwork; I _b' be the gray-scale value of known pixels point in channel B in sample block.

4. a kind of Images uniting restorative procedure based on partial structurtes feature according to claim 1 or 2 or 3, is characterized in that, described weight a ₁, a ₂, a ₃can arrange according to picture structure, when image texture characteristic to be repaired enriches, a ₁account for the largest percentage; When T-shaped partial structurtes abundant information such as turnings in image to be repaired, a ₂account for the largest percentage; When linear structure is enriched, a ₃account for the largest percentage.

5. a kind of Images uniting restorative procedure based on partial structurtes feature according to claim 4, is characterized in that, described when image texture characteristic to be repaired enriches, a ₁: a ₂: a ₃=3:1:1; In described image to be repaired during the T-shaped partial structurtes abundant information of corner, a ₁: a ₂: a ₃=1:3:1; Described when linear structure is enriched, a ₁: a ₂: a ₃=1:1:3.

6., based on an Images uniting repair system for partial structurtes feature, it is characterized in that, comprising:

Load module, for inputting an image to be repaired, is designated as u (x, y), and x, y are respectively the coordinate of image to be repaired on x, y direction;

Initialization module, for the degree of confidence initialization to all pixels in described image to be repaired; Specific as follows:

$C\left(p^{\′}\right)=0,\∀p^{\′}\∈\mathrm{\Ω},C\left(p^{\′}\right)=1,\∀p^{\′}\∈\mathrm{\Φ}$

In formula, p' is the pixel in described image to be repaired; The degree of confidence that C (p') is pixel p'; Ω is the affected area in described image to be repaired, and Φ is the non-affected area in described image to be repaired; for getting arbitrary value; refer to that the affected area in described image to be repaired gets any pixel; refer to that the non-affected area in described image to be repaired gets any pixel;

Formwork chooses module, for choosing the formwork of each pixel described respectively in described image damaged boundary to be repaired centered by each pixel;

Computing module, for the size according to the damaged degree priority of each pixel place formwork in image damaged boundary to be repaired described in following formulae discovery, and by damaged degree priority according to descending sort from big to small;

P(p)＝a ₁C(p)+a ₂D(p)+a ₃H(p)；

In formula, p is the pixel of described formwork center; The damaged degree priority that P (p) is pixel p; a ₁, a ₂, a ₃for weight, and a ₁, a ₂, a ₃and be 1;

H (p)=Kh+exp (-h); K is controling parameter, and value is 0.8; Exp (-h)=e ^-h, e ^-hfor taking e as the exponential function of the truth of a matter, e=2.71828; H is local metric function, h=| λ ₁-λ ₂| ², λ ₁and λ ₂be respectively the First Eigenvalue and the Second Eigenvalue of described formwork structure tensor expression formula, expression formula is:

${\mathrm{\λ}}_{1,2}=\frac{1}{2}(j_{11}+j_{22}\±\sqrt{{(j_{11}+j_{22})}^2+4j_{12}^2})$

Wherein, j ₁₁, j ₁₂, j ₂₂obtained by its structure tensor expression formula to be repaired, described its structure tensor expression formula to be repaired is:

$J_{\mathrm{\ρ}}=\left(\begin{array}{cc}{j}_{11}& j_{12}\\ j_{21}& j_{22}\end{array}\right)=G_{\mathrm{\ρ}}*\left(\begin{array}{cc}{\left(\frac{\∂u(x,y)}{\∂x}\right)}^2& \frac{\∂u(x,y)}{\∂x}\·\frac{\∂u(x,y)}{\∂y}\\ \frac{\∂u(x,y)}{\∂x}\·\frac{\∂u(x,y)}{\∂y}& {\left(\frac{\∂u(x,y)}{\∂y}\right)}^2\end{array}\right)$

Wherein, G _ρfor taking ρ as the Gaussian kernel function of parameter, ρ=1; refer to G _ρwith ask convolution algorithm; refer to G _ρwith ask convolution algorithm; refer to G _ρwith ask convolution algorithm;

C (p) is the degree of confidence of described formwork center pixel p; D (p) is the data item of described formwork center pixel p;

$C\left(p\right)=\frac{\underset{p^{\′}\∈{\mathrm{\Ψ}}_P\∩\mathrm{\Ω}}{\Σ}C\left(p^{\′}\right)}{\left|{\mathrm{\Ψ}}_p\right|}$

$D\left(p\right)=\frac{|\▿I_p^{\⊥}\·n_p|}{\mathrm{\α}}$

In formula, ψ _pfor the formwork that damaged boundary is put centered by pixel p, Φ is the non-affected area in described image to be repaired, ψ _pthe common factor of the non-affected area in ∩ Φ finger print plate and described image to be repaired, for the pixel in the common factor of the affected area in described formwork and described image to be repaired, i.e. the pixel of non-affected area in described formwork, for pixel degree of confidence, | Ψ _p| be the area of formwork, the quantity of pixel in finger print plate; for the direction of the isophote of pixel p, i.e. the vertical direction of gradient, n _pfor the unit direction vector of pixel p, α is normalized factor, and in gray level image, value is 255; $C\left(\hat{p}\right)=C\left(p^{\′}\right)=1,\∀p^{\′}\∈\mathrm{\Φ};$

Matching module, for finding the highest region of the formwork similarity maximum with described damaged degree priority in the non-affected area of described image to be repaired, as blocks and optimal matching blocks;

Replication module, for copying in the maximum formwork of described damaged degree priority by the pixel value in described blocks and optimal matching blocks;

Update module, for upgrade the maximum formwork of described damaged degree priority according to following formula damaged area in the degree of confidence of all pixels, and return computing module, until image to be repaired is repaired completely;

$\begin{array}{cc}C\left(\hat{q}\right)=C\left(p_m\right)& \∀\hat{q}\∈{\mathrm{\Ψ}}_{p_m}\∩\mathrm{\Ω};\end{array}$

In formula, p _mfor the pixel of the center of the maximum formwork of described damaged degree priority; C (p _m) be pixel p _mdegree of confidence, for the formwork that described damaged degree priority is maximum, Ω is the affected area in described image to be repaired, refer to the formwork that described damaged degree priority is maximum with the common factor of affected area Ω in described image to be repaired, the affected area in the formwork that namely described damaged degree priority is maximum; refer to that in the formwork that described damaged degree priority is maximum, get any pixel in affected area is for the pixel in damaged area in the formwork that described damaged degree priority is maximum, for pixel in damaged area in the formwork that described damaged degree priority is maximum degree of confidence.

7. a kind of Images uniting repair system based on partial structurtes feature according to claim 6, it is characterized in that, described matching module comprises:

Judging unit, for traveling through the sample block of all non-affected area in described image to be repaired, the size of the formwork that described sample block is maximum with described damaged degree priority is identical, and in judging the formwork that described damaged degree priority is maximum in all pixel value sums and described sample block all pixel value sums whether meet following relational expression:

(1-δ)·sum(Ψ _p”)≤sum(Ψ _q')≤(1+δ)·sum(Ψ _p”)；

In formula, p " be the known pixel of pixel in the maximum formwork of described damaged degree priority, Ψ _{p "}for the region that pixel in the formwork that described damaged degree priority is maximum is known; Sum (Ψ _{p "}) refer to that in the formwork maximum to described damaged degree priority, all pixel point values of pixel known portions are sued for peace; Q' is the corresponding pixel of pixel that in formwork maximum with described damaged degree priority in described sample block, pixel is known, Ψ _q'for the region that the region that pixel in formwork maximum with described damaged degree priority in described sample block is known is corresponding; Sum (Ψ _q') refer to that the pixel point value corresponding to known pixels point in formwork maximum with described damaged degree priority in described sample block is sued for peace; δ value is [0,1];

Color-match unit, for after all pixel value sums meet the relational expression in described judging unit in pixel value sums all in described sample block and the maximum formwork of the maximum described damaged degree priority of described damaged degree priority, carries out color-match;

Optimum Matching unit, for using sample block the highest for color-match degree as blocks and optimal matching blocks.

8. a kind of Images uniting repair system based on partial structurtes feature according to claim 7, it is characterized in that, described color-match unit carries out color-match according to following formula:

${\mathrm{\Ψ}}_{\tilde{q}}=\arg \underset{{\mathrm{\Ψ}}_q\⋐\mathrm{\Φ}}{\min }d({\mathrm{\Ψ}}_{p_m},{\mathrm{\Ψ}}_q)+\underset{i=1}{\overset{n}{\Σ}}\[{({\mathrm{\λ}}_1-{\mathrm{\λ}}_1^{\′})}^2+{({\mathrm{\λ}}_2-{\mathrm{\λ}}_2^{\′})}^2\]$

In formula, refer to ask for make get the ψ of minimum value _q; λ ₁for the First Eigenvalue of the maximum formwork structure tensor expression formula of described damaged degree priority; λ ₂for the Second Eigenvalue of the maximum formwork structure tensor expression formula of described damaged degree priority; λ ' ₁for the First Eigenvalue of sample block structure tensor expression formula; λ ' ₂the Second Eigenvalue of sample block structure tensor expression formula; Φ is non-affected area in described image to be repaired; for the formwork that described damaged degree priority is maximum; Q is the pixel of the formwork center that in sample block, corresponding described damaged degree priority is maximum; ψ _qfor the sample block put centered by pixel q; N is the quantity of all pixels in described formwork; for the sample block that color-match degree in non-affected area is the highest;

the formwork maximum for described damaged degree priority and the color sum of squares of deviations of sample block, expression formula is:

$d({\mathrm{\Ψ}}_{p_m},{\mathrm{\Ψ}}_q)=\mathrm{\Σ}\[{(I_R-I_R^{\′})}^2+{(I_G-I_G^{\′})}^2+{(I_B-I_B^{\′})}^2\]$

In formula, I _rfor the gray-scale value of known pixels point in R passage in formwork; I _r' be the gray-scale value of known pixels point in R passage in sample block; I _gfor the gray-scale value of known pixels point in G passage in formwork; I _g' be the gray-scale value of known pixels point in G passage in sample block; I _bfor the gray-scale value of known pixels point in channel B in formwork; I _b' be the gray-scale value of known pixels point in channel B in sample block.

9. a kind of Images uniting repair system based on partial structurtes feature according to claim 6 or 7 or 8, it is characterized in that, described computing module comprises setting unit, for arranging weight a according to picture structure ₁, a ₂, a ₃, when image texture characteristic to be repaired enriches, a ₁account for the largest percentage; When T-shaped partial structurtes abundant information such as turnings in image to be repaired, a ₂account for the largest percentage; When linear structure is enriched, a ₃account for the largest percentage.

10. a kind of Images uniting repair system based on partial structurtes feature according to claim 9, is characterized in that, described setting unit also for when image texture characteristic to be repaired enriches, a ₁: a ₂: a ₃=3:1:1; In described image to be repaired during the T-shaped partial structurtes abundant information of corner, a ₁: a ₂: a ₃=1:3:1; Described when linear structure is enriched, a ₁: a ₂: a ₃=1:1:3.