CN103258317B - The method realizing image color correction conversion based on sample image in computer system - Google Patents

Abstract

The present invention relates to a kind of method realizing image color correction conversion based on sample image in computer system, set up the matrixing model of rgb space including system, rgb space reads in sample image and target image, calculate the transformation matrix from sample image to target image according to the minimum chromatism method of Newton iteration, update the rgb space component of target image according to matrixing model by transformation matrix and complete the colour correction conversion of target image。Adopt the method realizing image color correction conversion based on sample image in this kind of computer system, overcome color of image problem pockety；It is simultaneously achieved the calculating of coupling matrix, the method optimized by newton aberration calculates target image colourity, overcome the tradition insurmountable problem of color correcting method based on sample image, ensure that the colour correction correctness in global and local, and processing procedure simple and fast, efficiency is higher, stable and reliable working performance, and the scope of application is relatively broad。

Description

The method realizing image color correction conversion based on sample image in computer system

Technical field

The present invention relates to Computer Image Processing field, particularly to image color correction technical field, a kind of method specifically referring to realize image color correction conversion based on sample image in computer system。

Background technology

Image color correction is the processing procedure changing input picture color。The colour correction technology of early stage is mainly used in photochrome or film, and generally adopts artificially colored method to process photo or film。Such workload done is very huge。Along with the appearance of computer technology, computer assisted color correcting method drastically increases the work efficiency of process。Computer assisted color correcting method is generally divided into: based on the method for the method of template and semi-artificial intervention。The former is obtain suitable image template at the ultimate challenge doing colour correction。The latter is then the color adjusting target image by adding manual intervention。Therefore relative to the color correcting method of manual intervention, based on template human input more less than the colour correction needs joined。

Through the literature search of prior art is found, the algorithm adopted based on the color correcting method of manual intervention is mainly by the dependency of brightness and colourity, neighbor similarity and geodesic distance are at interior multiple characteristics of image, such as A.Levin., D.Lischinski., Y.Weiss is at " ACMTransactionOnGraphics " (american computer association graphics transactions) the 23rd volume, 3rd phase, 689th page to 694 pages " ColorizationUsingOptimization " delivered literary compositions propose assume based on yuv space brightness and chroma dependency, by the method that Minimum Mean Square Error solves target image colourity。ReinhardE, AdhikhminM is at " IEEEComputerGraphicsandApplication " (IEEE computer graphical and application) the 21st volume, 5th phase, the 34th page to 41 pages " colortransferbetweenimages " delivered (color based on color space shifts) literary compositions propose the color correcting method based on coupling matrix。

Said method is all based on the method for global statistics information, make use of the dependency of color space to solve the colourity of target image, often adopts simple global statistics parameter such as average, variance etc. when color notation conversion space。But, these methods fail to consider the heterochromia of image local, thus in the discordance of local distribution after being easily caused colour correction,。

Summary of the invention

It is an object of the invention to overcome above-mentioned shortcoming of the prior art, it is provided that a kind of method that ensure that colour correction realizes image color correction conversion based on sample image in the correctness of global and local, processing procedure simple and fast, efficiency is higher, stable and reliable working performance, the scope of application are relatively broad computer system。

In order to realize above-mentioned purpose, in the computer system of the present invention based on sample image realize image color correction conversion method as follows:

The method realizing image color correction conversion based on sample image in this computer system, it is mainly characterized by, and described method comprises the following steps:

(1) system sets up the matrixing model of rgb space；

(2) system reads in sample image and target image in described rgb space；

(3) system calculates the transformation matrix from described sample image to target image according to the minimum chromatism method of Newton iteration；

(4) according to described matrixing model, the rgb space component of described target image is updated by described transformation matrix, thus completing the colour correction conversion of target image。

This computer system realizes the matrixing model setting up rgb space in the method for image color correction conversion based on sample image, particularly as follows:

According to below equation, rgb space is set up transformation matrix:

Y=M x；

Wherein, $x=\left[\begin{array}{cccc}{R}_1^{*}& R_2^{*}& & R_N^{*}\\ G_1^{*}& G_2^{*}& \mathrm{...}& G_N^{*}\\ B_1^{*}& B_2^{*}& & B_N^{*}\end{array}\right]$ For sample image matrix, $\left[\begin{array}{cccc}{R}_1^{*}& R_2^{*}& \mathrm{...}& R_N^{*}\end{array}\right]$ For the R component value of sample image, $\left[\begin{array}{cccc}{G}_1^{*}& G_2^{*}& \mathrm{...}& G_N^{*}\end{array}\right]$ For the G component value of sample image, $\left[\begin{array}{cccc}{B}_1^{*}& B_2^{*}& \mathrm{...}& B_N^{*}\end{array}\right]$ For the B component value of sample image, N is image pixel index number, $y=\left[\begin{array}{cccc}{R}_1& R_2& & R_N\\ G_1& G_2& \mathrm{...}& G_N\\ B_1& B_2& & B_N\end{array}\right]$ For target image matrix, [R₁R₂…R_N] for the R component value of target image, [G₁G₂…G_N] for the G component value of target image, [B₁B₂…B_N] for the B component value of target image, $M=\left[\begin{array}{ccc}{m}_{11}& m_{12}& m_{13}\\ m_{21}& m_{22}& m_{23}\\ m_{31}& m_{32}& m_{33}\end{array}\right]$ For transformation matrix, m₁₁,m₁₂,m₁₃For image R component conversion coefficient, m₂₁,m₂₂,m₂₃For image G component conversion coefficient, m₃₁,m₃₂,m₃₃For image B component conversion coefficient。

This computer system realizes based on sample image the method for image color correction conversion calculates transformation matrix from described sample image to target image according to the minimum chromatism method of Newton iteration, comprises the following steps:

(31) rectangular coordinate system is set up with the pixel in the upper left corner in described sample image for initial point；

(32) iteration primary quantity T is calculated according to below equation₁、T₂、T₃:

$T_1=X^T\·M_1=\left[\begin{array}{c}X{(R,G,B)}_{\left(1\right)1}\\ X{(R,G,B)}_{\left(1\right)2}\\ .\\ .\\ .\\ X{(R,G,B)}_{\left(1\right)N}\end{array}\right]=\left[\begin{array}{c}{R}_1\\ R_2\\ .\\ .\\ .\\ R_N\end{array}\right]=\left[\begin{array}{ccc}{R}_1^{*}& G_1^{*}& B_1^{*}\\ R_2^{*}& G_2^{*}& B_2^{*}\\ & .& \\ & .& \\ & .& \\ R_N^{*}& G_N^{*}& B_N^{*}\end{array}\right]\·\left[\begin{array}{c}{m}_{11}\\ m_{12}\\ m_{13}\end{array}\right];$

$T_2=X^T\·M_2=\left[\begin{array}{c}X{(R,G,B)}_{\left(2\right)1}\\ X{(R,G,B)}_{\left(2\right)2}\\ .\\ .\\ .\\ X{(R,G,B)}_{\left(2\right)N}\end{array}\right]=\left[\begin{array}{c}{G}_1\\ G_2\\ .\\ .\\ .\\ G_N\end{array}\right]=\left[\begin{array}{ccc}{R}_1^{*}& G_1^{*}& B_1^{*}\\ R_2^{*}& G_2^{*}& B_2^{*}\\ & .& \\ & .& \\ & .& \\ R_N^{*}& G_N^{*}& B_N^{*}\end{array}\right]\·\left[\begin{array}{c}{m}_{21}\\ m_{22}\\ m_{23}\end{array}\right];$

$T_3=X^T\·M_3=\left[\begin{array}{c}X{(R,G,B)}_{\left(3\right)1}\\ X{(R,G,B)}_{\left(3\right)2}\\ .\\ .\\ .\\ X{(R,G,B)}_{\left(3\right)N}\end{array}\right]=\left[\begin{array}{c}{B}_1\\ B_2\\ .\\ .\\ .\\ B_N\end{array}\right]=\left[\begin{array}{ccc}{R}_1^{*}& G_1^{*}& B_1^{*}\\ R_2^{*}& G_2^{*}& B_2^{*}\\ & .& \\ & .& \\ & .& \\ R_N^{*}& G_N^{*}& B_N^{*}\end{array}\right]\·\left[\begin{array}{c}{m}_{31}\\ m_{32}\\ m_{33}\end{array}\right];$

Wherein, X (R, G, B)₍₁₎₁,...,X(R,G,B)_(1)NFor the R component of target image, X (R, G, B)₍₂₎₁,...,X(R,G,B)_(2)NFor the G component of target image, X (R, G, B)₍₃₎₁,...,X(R,G,B)_(3)NFor the B component of target image, T is for turning order operator, and N is pixel pixel index value in rgb space, M₁For image R component conversion coefficient vector, M₂For image G component conversion coefficient vector, M₃For image B component conversion coefficient vector；

(33) it is iterated calculating according to following iteration more new-standard cement:

T_(k)i+1=T_(k)i+Δ_(k)i；

Wherein, i is iterations, Δ_(k)i=-J_(k)iε_(k)i, Δ_(k)iIt is the color component residual error in the i-th step iteration in kth spatial component, J_(k)iIt is the Jacobian matrix of 3 × N in kth spatial component, ε in the i-th step iteration_(k)iIt is the color space residual error in the i-th step iteration in kth spatial component, and:

ε_(k)i=x^T·M_(k)i-T_(k)i, k=1,2,3；

Wherein, k is rgb color space component index value, and k=1 represents R space, and k=2 represents G space, and k=3 represents B space, and i is iterations, and x is sample image matrix, wherein comprises RGB component value, M_(k)iIt is the vector in kth spatial component of the transformation matrix M in the i-th step iteration, T_(k)iIt it is the iteration updated value in the i-th step iteration in kth spatial component；

(34) whether the iterative computation of real-time judge step (33) has met relationship below:

||ε_(k)i+J_(k)iΔ_(k)i| | < ε '

Wherein, | | | | for modulo operation, ε ' is iteration threshold, and ε is color space residual error, and J is the Jacobian matrix of 3 × N, and Δ is that target image color component iteration is poor；

(35) if be unsatisfactory for, then step (33) is continued；

(36) if it is satisfied, then obtain corresponding optimum color transformed matrix M, and $M=\left[\begin{array}{ccc}{m}_{11}& m_{12}& m_{13}\\ m_{21}& m_{22}& m_{23}\\ m_{31}& m_{32}& m_{33}\end{array}\right].$

The iteration threshold ε ' realized based on sample image in this computer system in the method for image color correction conversion is 0.6。

Have employed the method realizing image color correction conversion based on sample image in the computer system of this invention, owing to wherein make use of what coupling matrixing and the optimization method of Newton iteration method established target image color space to solve framework, thus overcoming color of image problem pockety；Fully taken into account the diversity in color transformation process simultaneously, by adopting the optimization method based on Newton iteration method to achieve the calculating of coupling matrix, owing to the method that colour correction is optimized by newton aberration is calculated target image colourity, thus overcoming the tradition insurmountable problem of color correcting method based on sample image, ensure that the colour correction correctness in global and local, and processing procedure simple and fast, efficiency is higher, stable and reliable working performance, the scope of application is relatively broad。

Accompanying drawing explanation

Fig. 1 be the present invention computer system in realize the overall flow figure of method of image color correction conversion based on sample image。

Fig. 2 a be the present invention computer system in based on sample image realize image color correction conversion method specific embodiment in sample image。

Fig. 2 b is the image RGB rectangular histogram of Fig. 2 a。

Fig. 3 a be the present invention computer system in based on sample image realize image color correction conversion method specific embodiment in target image。

Fig. 3 b is the image RGB rectangular histogram of Fig. 3 a。

Fig. 4 a be the present invention computer system in based on sample image realize image color correction conversion method specific embodiment in treatment effect image。

Fig. 4 b is the image RGB rectangular histogram of Fig. 4 a。

Detailed description of the invention

In order to be more clearly understood that the technology contents of the present invention, describe in detail especially exemplified by following example。

Refer to shown in Fig. 1 to Fig. 4 b, the method realizing image color correction conversion based on sample image in this computer system, including following steps:

(1) system sets up the matrixing model of rgb space, particularly as follows:

According to below equation, rgb space is set up transformation matrix:

Y=M x；

Wherein, $x=\left[\begin{array}{cccc}{R}_1^{*}& R_2^{*}& & R_N^{*}\\ G_1^{*}& G_2^{*}& \mathrm{...}& G_N^{*}\\ B_1^{*}& B_2^{*}& & B_N^{*}\end{array}\right]$ For sample image matrix, $\left[\begin{array}{cccc}{R}_1^{*}& R_2^{*}& \mathrm{...}& R_N^{*}\end{array}\right]$ For the R component value of sample image, $\left[\begin{array}{cccc}{G}_1^{*}& G_2^{*}& \mathrm{...}& G_N^{*}\end{array}\right]$ For the G component value of sample image, $\left[\begin{array}{cccc}{B}_1^{*}& B_2^{*}& \mathrm{...}& B_N^{*}\end{array}\right]$ For the B component value of sample image, N is image pixel index number, $y=\left[\begin{array}{cccc}{R}_1& R_2& & R_N\\ G_1& G_2& \mathrm{...}& G_N\\ B_1& B_2& & B_N\end{array}\right]$ For target image matrix, [R₁R₂…R_N] for the R component value of target image, [G₁G₂…G_N] for the G component value of target image, [B₁B₂…B_N] for the B component value of target image, $M=\left[\begin{array}{ccc}{m}_{11}& m_{12}& m_{13}\\ m_{21}& m_{22}& m_{23}\\ m_{31}& m_{32}& m_{33}\end{array}\right]$ For transformation matrix, m₁₁,m₁₂,m₁₃For image R component conversion coefficient, m₂₁,m₂₂,m₂₃For image G component conversion coefficient, m₃₁,m₃₂,m₃₃For image B component conversion coefficient；

(2) system reads in sample image and target image in described rgb space；

(3) system calculates the transformation matrix from described sample image to target image according to the minimum chromatism method of Newton iteration, comprises the following steps:

A () sets up rectangular coordinate system with the pixel in the upper left corner in described sample image for initial point；

B () calculates iteration primary quantity T according to below equation₁、T₂、T₃:

$T_1=X^T\&CenterDot;M_1=\left[\begin{array}{c}X{(R,G,B)}_{\left(1\right)1}\\ X{(R,G,B)}_{\left(1\right)2}\\ .\\ .\\ .\\ X{(R,G,B)}_{\left(1\right)N}\end{array}\right]=\left[\begin{array}{c}{R}_1\\ R_2\\ .\\ .\\ .\\ R_N\end{array}\right]=\left[\begin{array}{ccc}{R}_1^{*}& G_1^{*}& B_1^{*}\\ R_2^{*}& G_2^{*}& B_2^{*}\\ & .& \\ & .& \\ & .& \\ R_N^{*}& G_N^{*}& B_N^{*}\end{array}\right]\&CenterDot;\left[\begin{array}{c}{m}_{11}\\ m_{12}\\ m_{13}\end{array}\right];$

$T_2=X^T\&CenterDot;M_2=\left[\begin{array}{c}X{(R,G,B)}_{\left(2\right)1}\\ X{(R,G,B)}_{\left(2\right)2}\\ .\\ .\\ .\\ X{(R,G,B)}_{\left(2\right)N}\end{array}\right]=\left[\begin{array}{c}{G}_1\\ G_2\\ .\\ .\\ .\\ G_N\end{array}\right]=\left[\begin{array}{ccc}{R}_1^{*}& G_1^{*}& B_1^{*}\\ R_2^{*}& G_2^{*}& B_2^{*}\\ & .& \\ & .& \\ & .& \\ R_N^{*}& G_N^{*}& B_N^{*}\end{array}\right]\&CenterDot;\left[\begin{array}{c}{m}_{21}\\ m_{22}\\ m_{23}\end{array}\right];$

$T_3=X^T\&CenterDot;M_3=\left[\begin{array}{c}X{(R,G,B)}_{\left(3\right)1}\\ X{(R,G,B)}_{\left(3\right)2}\\ .\\ .\\ .\\ X{(R,G,B)}_{\left(3\right)N}\end{array}\right]=\left[\begin{array}{c}{B}_1\\ B_2\\ .\\ .\\ .\\ B_N\end{array}\right]=\left[\begin{array}{ccc}{R}_1^{*}& G_1^{*}& B_1^{*}\\ R_2^{*}& G_2^{*}& B_2^{*}\\ & .& \\ & .& \\ & .& \\ R_N^{*}& G_N^{*}& B_N^{*}\end{array}\right]\&CenterDot;\left[\begin{array}{c}{m}_{31}\\ m_{32}\\ m_{33}\end{array}\right];$

Wherein, X (R, G, B)₍₁₎₁,...,X(R,G,B)_(1)NFor the R component of target image, X (R, G, B)₍₂₎₁,...,X(R,G,B)_(2)NFor the G component of target image, X (R, G, B)₍₃₎₁,...,X(R,G,B)_(3)NFor the B component of target image, T is for turning order operator, and N is pixel pixel index value in rgb space, M₁For image R component conversion coefficient vector, M₂For image G component conversion coefficient vector, M₃For image B component conversion coefficient vector；

C () is iterated calculating according to following iteration more new-standard cement:

T_(k)i+1=T_(k)i+Δ_(k)i；

Wherein, i is iterations, Δ_(k)i=-J_(k)iε_(k)i, Δ_(k)iIt is the color component residual error in the i-th step iteration in kth spatial component, J_(k)iIt is the Jacobian matrix of 3 × N in kth spatial component, ε in the i-th step iteration_(k)iIt is the color space residual error in the i-th step iteration in kth spatial component, and:

ε_(k)i=x^T·M_(k)i-T_(k)i, k=1,2,3；

Wherein, k is rgb color space component index value, and k=1 represents R space, and k=2 represents G space, and k=3 represents B space, and i is iterations, and x is sample image matrix, wherein comprises RGB component value, M_(k)iIt is the vector in kth spatial component of the transformation matrix M in the i-th step iteration, T_(k)iIt it is the iteration updated value in the i-th step iteration in kth spatial component；

Whether the iterative computation of (d) real-time judge step (33) has met relationship below:

||ε_(k)i+J_(k)iΔ_(k)i| | < ε '

Wherein, | | | | for modulo operation, ε ' is iteration threshold, and ε is color space residual error, and J is the Jacobian matrix of 3 × N, and Δ is that target image color component iteration is poor；

If e () is unsatisfactory for, then continue step (c)；

(f) if it is satisfied, then obtain corresponding optimum color transformed matrix M, and $M=\left[\begin{array}{ccc}{m}_{11}& m_{12}& m_{13}\\ m_{21}& m_{22}& m_{23}\\ m_{31}& m_{32}& m_{33}\end{array}\right];$

(4) according to described matrixing model, the rgb space component of described target image is updated by described transformation matrix, thus completing the colour correction conversion of target image。

In the middle of actually used, the present invention is achieved by the following technical solutions, specifically includes following steps:

Step one, sets up transformation matrix at rgb space,

Y=M x ... (1)

X is sample image, and y is target image, and M is transformation matrix。Being described in detail below at rgb space:

$\left[\begin{array}{cccc}{R}_1& R_2& & R_N\\ G_1& G_2& \mathrm{...}& G_N\\ B_1& B_2& & B_N\end{array}\right]=\left[\begin{array}{ccc}{m}_{11}& m_{12}& m_{13}\\ m_{21}& m_{22}& m_{23}\\ m_{31}& m_{32}& m_{33}\end{array}\right]\&CenterDot;\left[\begin{array}{cccc}{R}_1^{*}& R_2^{*}& & R_N^{*}\\ G_1^{*}& G_2^{*}& \mathrm{...}& G_N^{*}\\ B_1^{*}& B_2^{*}& & B_N^{*}\end{array}\right]\mathrm{......}\left(2\right)$

Step 2, adopts the method for the minimum aberration of Newton iteration to calculate transformation matrix:

$T_1=X^T\&CenterDot;M_1=\left[\begin{array}{c}X{(R,G,B)}_{\left(1\right)1}\\ X{(R,G,B)}_{\left(1\right)2}\\ .\\ .\\ .\\ X{(R,G,B)}_{\left(1\right)N}\end{array}\right]=\left[\begin{array}{c}{R}_1\\ R_2\\ .\\ .\\ .\\ R_N\end{array}\right]=\left[\begin{array}{ccc}{R}_1^{*}& G_1^{*}& B_1^{*}\\ R_2^{*}& G_2^{*}& B_2^{*}\\ & .& \\ & .& \\ & .& \\ R_N^{*}& G_N^{*}& B_N^{*}\end{array}\right]\&CenterDot;\left[\begin{array}{c}{m}_{11}\\ m_{12}\\ m_{13}\end{array}\right]\mathrm{......}\left(3\right)$

$T_2=X^T\&CenterDot;M_2=\left[\begin{array}{c}X{(R,G,B)}_{\left(2\right)1}\\ X{(R,G,B)}_{\left(2\right)2}\\ .\\ .\\ .\\ X{(R,G,B)}_{\left(2\right)N}\end{array}\right]=\left[\begin{array}{c}{G}_1\\ G_2\\ .\\ .\\ .\\ G_N\end{array}\right]=\left[\begin{array}{ccc}{R}_1^{*}& G_1^{*}& B_1^{*}\\ R_2^{*}& G_2^{*}& B_2^{*}\\ & .& \\ & .& \\ & .& \\ R_N^{*}& G_N^{*}& B_N^{*}\end{array}\right]\&CenterDot;\left[\begin{array}{c}{m}_{21}\\ m_{22}\\ m_{23}\end{array}\right]\mathrm{......}\left(4\right)$

$T_3=X^T\&CenterDot;M_3=\left[\begin{array}{c}X{(R,G,B)}_{\left(3\right)1}\\ X{(R,G,B)}_{\left(3\right)2}\\ .\\ .\\ .\\ X{(R,G,B)}_{\left(3\right)N}\end{array}\right]=\left[\begin{array}{c}{B}_1\\ B_2\\ .\\ .\\ .\\ B_N\end{array}\right]=\left[\begin{array}{ccc}{R}_1^{*}& G_1^{*}& B_1^{*}\\ R_2^{*}& G_2^{*}& B_2^{*}\\ & .& \\ & .& \\ & .& \\ R_N^{*}& G_N^{*}& B_N^{*}\end{array}\right]\&CenterDot;\left[\begin{array}{c}{m}_{31}\\ m_{32}\\ m_{33}\end{array}\right]\mathrm{......}\left(5\right)$

Assumed appearance color space residual error is:

ε_(k)i=x^T·M_(k)i-T_(k)i, k=1,2,3 ... (6)

I represents iterations, and iteration more new-standard cement is:

T_(k)i+1=T_(k)i+Δ_(k)i……(7)

Δ_(K)i=-J_(k)iε_(k)i……(8)

J is the Jacobian matrix of 3 × N, updates termination and meets following formula:

||ε_(k)i+J_(k)iΔ_(k)i| | < ε ' ... (9)

When constraints meets, we obtain optimum color transformed matrix M from sample image point and target image point。

Step 2, obtains the final three-channel color component of target image RGB by the transformation matrix of step 2。

Compared with prior art, the invention has the beneficial effects as follows: utilize what coupling matrixing and the optimization method of Newton iteration method set up target image color space to solve framework, overcome color of image problem pockety；The color correcting method of existing sample image, simple global statistics parameter such as average, variance etc. is often adopted when color notation conversion space, and the present invention has fully taken into account the diversity in color transformation process, by adopting the optimization method based on Newton iteration method to achieve the calculating of coupling matrix。Owing to the method that colour correction is optimized by newton aberration is calculated target image colourity by the present invention, thus overcoming the tradition insurmountable problem of color correcting method based on sample image。

In inventive embodiment, the image that 1 image is sized to 256 × 256 pixels does colour correction, and algorithm flow refers to shown in Fig. 1, comprises the steps:

The first step, reads in sample image and target image。

Second step, calculates the sample image transformation matrix to target image, and specific formula for calculation is as follows:

$T_1=X^T\&CenterDot;M_1=\left[\begin{array}{c}X{(R,G,B)}_{\left(1\right)1}\\ X{(R,G,B)}_{\left(1\right)2}\\ .\\ .\\ .\\ X{(R,G,B)}_{\left(1\right)N}\end{array}\right]=\left[\begin{array}{c}{R}_1\\ R_2\\ .\\ .\\ .\\ R_N\end{array}\right]=\left[\begin{array}{ccc}{R}_1^{*}& G_1^{*}& B_1^{*}\\ R_2^{*}& G_2^{*}& B_2^{*}\\ & .& \\ & .& \\ & .& \\ R_N^{*}& G_N^{*}& B_N^{*}\end{array}\right]\&CenterDot;\left[\begin{array}{c}{m}_{11}\\ m_{12}\\ m_{13}\end{array}\right];$

$T_2=X^T\&CenterDot;M_2=\left[\begin{array}{c}X{(R,G,B)}_{\left(2\right)1}\\ X{(R,G,B)}_{\left(2\right)2}\\ .\\ .\\ .\\ X{(R,G,B)}_{\left(2\right)N}\end{array}\right]=\left[\begin{array}{c}{G}_1\\ G_2\\ .\\ .\\ .\\ G_N\end{array}\right]=\left[\begin{array}{ccc}{R}_1^{*}& G_1^{*}& B_1^{*}\\ R_2^{*}& G_2^{*}& B_2^{*}\\ & .& \\ & .& \\ & .& \\ R_N^{*}& G_N^{*}& B_N^{*}\end{array}\right]\&CenterDot;\left[\begin{array}{c}{m}_{21}\\ m_{22}\\ m_{23}\end{array}\right];$

$T_3=X^T\&CenterDot;M_3=\left[\begin{array}{c}X{(R,G,B)}_{\left(3\right)1}\\ X{(R,G,B)}_{\left(3\right)2}\\ .\\ .\\ .\\ X{(R,G,B)}_{\left(3\right)N}\end{array}\right]=\left[\begin{array}{c}{B}_1\\ B_2\\ .\\ .\\ .\\ B_N\end{array}\right]=\left[\begin{array}{ccc}{R}_1^{*}& G_1^{*}& B_1^{*}\\ R_2^{*}& G_2^{*}& B_2^{*}\\ & .& \\ & .& \\ & .& \\ R_N^{*}& G_N^{*}& B_N^{*}\end{array}\right]\&CenterDot;\left[\begin{array}{c}{m}_{31}\\ m_{32}\\ m_{33}\end{array}\right];$

Assumed appearance color space residual error is:

ε_(k)i=x^T·M_(k)i-T_(k)i, k=1,2,3；

I represents iterations, and iteration more new-standard cement is:

T_(k)i+1=T_(k)i+Δ_(k)i；

Δ_(K)i=-J_(k)iε_(k)i；

J is the Jacobian matrix of 3 × N, updates termination and meets following formula:

||ε_(k)i+J_(k)iΔ_(k)i| | < ε '；

Threshold epsilon ' generally value be 0.6, so, we obtain the color transformed matrix M of the point from sample image to target image。

3rd step, is updated the rgb space of target image, uses formula calculated as below by transformation matrix:

$\left[\begin{array}{cccc}{R}_1& R_2& & R_N\\ G_1& G_2& \mathrm{...}& G_N\\ B_1& B_2& & B_N\end{array}\right]=\left[\begin{array}{ccc}{m}_{11}& m_{12}& m_{13}\\ m_{21}& m_{22}& m_{23}\\ m_{31}& m_{32}& m_{33}\end{array}\right]\&CenterDot;\left[\begin{array}{cccc}{R}_1^{*}& R_2^{*}& & R_N^{*}\\ G_1^{*}& G_2^{*}& \mathrm{...}& G_N^{*}\\ B_1^{*}& B_2^{*}& & B_N^{*}\end{array}\right].$

The principle of the colour correction of the present embodiment is: by the calculating of color transformed matrix, introduces newton residual minimization method, efficiently solves the problem that distribution of color is uneven；Under Unified frame, achieve the colour correction of image, obtain target image color space values by Optimized Matching matrix calculus；Therefore, by establishing a kind of Computational frame based on coupling matrix, it may be achieved the colour correction of effective image。

For Fig. 2 b, 3b and 4b, it corresponds respectively to the RGB rectangular histogram of Fig. 2 a, 3a, 4a, and wherein horizontal axis is gray value coordinate, and vertical axis represents the quantity of pixel under given gray value。

Have employed the method realizing image color correction conversion based on sample image in above-mentioned computer system, owing to wherein make use of what coupling matrixing and the optimization method of Newton iteration method established target image color space to solve framework, thus overcoming color of image problem pockety；Fully taken into account the diversity in color transformation process simultaneously, by adopting the optimization method based on Newton iteration method to achieve the calculating of coupling matrix, owing to the method that colour correction is optimized by newton aberration is calculated target image colourity, thus overcoming the tradition insurmountable problem of color correcting method based on sample image, ensure that the colour correction correctness in global and local, and processing procedure simple and fast, efficiency is higher, stable and reliable working performance, the scope of application is relatively broad。

In this description, the present invention is described with reference to its specific embodiment。But it is clear that still may be made that various amendment and conversion are without departing from the spirit and scope of the present invention。Therefore, specification and drawings is regarded in an illustrative, rather than a restrictive。

Claims (2)

1. the method realizing image color correction conversion based on sample image in a computer system, it is characterised in that described method comprises the following steps:

(1) system sets up the matrixing model of rgb space；

(2) system reads in sample image and target image in described rgb space；

(3) system calculates the transformation matrix from described sample image to target image according to the minimum chromatism method of Newton iteration；

(4) according to described matrixing model, the rgb space component of described target image is updated by described transformation matrix, thus completing the colour correction conversion of target image；

The described matrixing model setting up rgb space, particularly as follows:

According to below equation, rgb space is set up transformation matrix:

Y=M x；

Wherein,For sample image matrix,For the R component value of sample image,For the G component value of sample image,For the B component value of sample image, N is image pixel index number,For target image matrix, [R₁R₂…R_N] for the R component value of target image, [G₁G₂…G_N] for the G component value of target image, [B₁B₂…B_N] for the B component value of target image,For transformation matrix, m₁₁,m₁₂,m₁₃For image R component conversion coefficient, m₂₁,m₂₂,m₂₃For image G component conversion coefficient, m₃₁,m₃₂,m₃₃For image B component conversion coefficient；

Described calculates the transformation matrix from described sample image to target image according to the minimum chromatism method of Newton iteration, comprises the following steps:

(31) rectangular coordinate system is set up with the pixel in the upper left corner in described sample image for initial point；

(32) iteration primary quantity T is calculated according to below equation₁、T₂、T₃:

Wherein, X is the rgb space vector of target image, X (R, G, B)₍₁₎₁,...,X(R,G,B)_(1)NFor the R component of target image, X (R, G, B)₍₂₎₁,...,X(R,G,B)_(2)NFor the G component of target image, X (R, G, B)₍₃₎₁,...,X(R,G,B)_(3)NFor the B component of target image, T is for turning order operator, and N is pixel pixel index value in rgb space, M₁For image R component conversion coefficient vector, M₂For image G component conversion coefficient vector, M₃For image B component conversion coefficient vector；

(33) it is iterated calculating according to following iteration more new-standard cement:

T_(k)i+1=T_(k)i+Δ_(k)i；

Wherein, i is iterations, Δ_(k)i=-J_(k)iε_(k)i, Δ_(k)iIt is the color component residual error in the i-th step iteration in kth spatial component, J_(k)iIt is the Jacobian matrix of 3 × N in kth spatial component, ε in the i-th step iteration_(k)iIt is the color space residual error in the i-th step iteration in kth spatial component, and:

ε_(k)i=x^T·M_(k)i-T_(k)i, k=1,2,3；

Wherein, k is rgb color space component index value, and k=1 represents R space, and k=2 represents G space, and k=3 represents B space, and i is iterations, and x is sample image matrix, wherein comprises RGB component value, M_(k)iIt is the vector in kth spatial component of the transformation matrix M in the i-th step iteration, T_(k)iIt it is the iteration updated value in the i-th step iteration in kth spatial component；

(34) whether the iterative computation of real-time judge step (33) has met relationship below:

||ε_(k)i+J_(k)iΔ_(k)i| | < ε '

Wherein, | | | | for modulo operation, ε ' is iteration threshold, and ε is color space residual error, and J is the Jacobian matrix of 3 × N, and Δ is that target image color component iteration is poor；

(35) if be unsatisfactory for, then step (33) is continued；

(36) if it is satisfied, then obtain corresponding optimum color transformed matrix M, and

2. the method realizing image color correction conversion based on sample image in computer system according to claim 1, it is characterised in that described iteration threshold ε ' is 0.6。