CN105957111A - Sequence remote sensing image tone consistency correction method and system - Google Patents

Abstract

The invention discloses a sequence remote sensing image tone consistency correction method and a system. The method comprises the following steps: transferring a colored sequence image from RGB color space to L[alpha, beta] color space to perform the tone transfer preprocessing and the global gain compensation of luminance channels; in color channel histogram mapping, adjusting the image gray-scales of color channel [alpha] and color channel [beta] separately; transferring all images from L[alpha, beta] color space to RGB color space and outputting and saving the processed results. According to the invention, the method and system can effectively suppress or eliminate the differences in brightness and the color existing in the images by utilizing the tone corresponding relationship of the overlapping areas of adjacent images. The method and system can also greatly reduce the image distortion caused by the correlation between the channels, and automatically select the reference color by searching the maximum consistent image subset, thus avoiding the problem of manually selecting the reference color.

Description

The color tone consistency bearing calibration of sequence remote sensing image and system

Technical field

The invention belongs to Remote Sensing Image Processing Technology field, particularly relate to sequence remote sensing image color tone consistency bearing calibration and System.

Background technology

Remote sensing image is in an increasingly wide range of applications at aspects such as geographic mapping, disaster monitoring, resource investigation and urban plannings. The ring in flow process is quickly spliced, for comparing other modules, in terms of the light and color homogenization process of sequential images as remote sensing image Achievement in research is less, and is concentrated mainly on the dodging in terms of image brilliance, specifically includes that the even smooth algorithm of MASK, Wallis Filter even smooth algorithm etc..The most classical and the process of main flow is compared for color distortion problems such as the colour cast aberration of existence between image Method.For large-scale measured zone, due to the impact of the aspect such as lighting angle and cloud cover, the brightness between image and Color distortion is difficult to avoid, and this is all the more so for the remote sensing image of multidate.The light and color homogenization of sequence remote sensing image processes and relates to And to several key technologies: chromatic image luminance channel and the isolation technics of Color Channel, the tone Transfer Technology of image, sequence The optimum consistent color of image is with reference to subset search technology；

Brightness and the isolation technics of Color Channel: the color space of presently described color has a lot, show for equipment is exactly warp The rgb space of allusion quotation.Owing to there is bigger dependency between tri-passages of RBG, this image processing algorithms band to subchannel It is inconvenient to come, and therefore a lot of image algorithms are existing forwards to the space that Inter-channel Correlation is the lowest, go back to RGB after having processed again by image Space.It is that the separate color space of passage mainly has HSV space and L α β space by chromatic image by luminance channel and color.Its In, in L α β space, each interchannel dependency is minimum, and L * channel represents that luminance channel, α and β passage represent two orthogonal face Chrominance channel, is the color space of a preferable image processing.

The tone Transfer Technology of image: the tone Transfer Technology of image refers to pass to pending by the tone characteristics with reference to image Image, makes pending image have the tone with reference to image on the premise of keeping original presentation content.Main method has two classes: Conversion based on gray-scale statistical parameter (average and variance) and the conversion mapped based on grey level histogram.

There is presently no disclosed consistent color reference subset search based on sequential images theoretical.The final purpose of light and color homogenization is Spliced remote sensing image is allowed to present unified tone.When sequence remote sensing image exists multiple inconsistent tone, it is difficult to Determine that any is optimal standard colour tone.

Summary of the invention

The present invention is directed to the tone inconsistence problems that sequence remote sensing image is likely to occur when splicing, propose the color of sequence remote sensing image Adjust Concordance method and system.

The present invention provides the color tone consistency bearing calibration of a kind of sequence remote sensing image, comprises the following steps:

Step 1, forwards colored sequential images to L α β color space from RGB color；

Step 2, tone transmission pretreatment, including based on reference to image, utilize the Wallis Filtering Formula institute to image of cum rights Pixel is had to carry out tone transmission respectively at passage L α β；

Step 3, the global gain of luminance channel compensates, adjusts image greyscale in luminance channel L including performing following sub-step It is whole,

Step 3.1, adds up all adjacent images to the gray average μ (I in overlapped public territory_ij), wherein, I_ijRepresent image i In with the lap of image j；

Step 3.2, according to the energy of gain compensation, optimizes the gray scale gain coefficient a of every image_i, i ∈ [1, N], wherein N is image number Amount；

Step 3.3, according to the gain compensation factor that step 3.2 gained is optimum, does gain compensation to the luminance channel of every image as follows,

${\hat{I}}_i\left(p\right)=a_i{I}_i\left(p\right)$

Wherein, I_i(p) andRepresent arbitrary pixel p gray value before and after gain compensation on image i respectively；

Step 4, the Histogram Mapping of Color Channel, including performing following sub-step at Color Channel α and β separately to shadow As gray scale adjusts,

Step 4.1, builds undirected weighted graph, including individual image is considered as a node, will have between the image node of overlapping region Being connected with limit, limit power is set to this image to gray difference value D in overlapping region_H；

Step 4.2, sets threshold value T_D, work as D_H(I_i,I_j)≤T_D, then I_iAnd I_jLimit between image node is the consistent limit of tone, otherwise, For the non-uniform limit of tone；

Step 4.3, the reference image set that search is optimum, it is included in undirected weighted graph search and there is the largest connected son on the consistent limit of tone Figure, the reference image set mapped as color channel histograms using the image node set of this subgraph；

Step 4.4, determines the directly reference image of every image, sets including by the weights between the reference image node in undirected weighted graph It is 0, appoints and take one of them node as root node；Shortest path first is utilized to determine the shortest to root node of each image node Path, obtains binary tree, and the father node of each image node is exactly that it carries out the directly reference image of grey level histogram mapping；

Step 4.5, determines the grey scale mapping order of image sequence, including to the binary tree obtained in step 4.4, presses range from root node Preferential or the mode binary tree traversal of depth-first, the image node sequence of actual access be just by as pair between grey scale mapping Execution sequence；

Step 4.6, according to the execution sequence searched out, every image is set up the Histogram Mapping relation between its reference image, and is adjusted The gray scale of each pixel in whole image；

All images are gone to RGB color from L α β color space, export and preserve result by step 5.

And, adjacent image between the calculating process of gray difference value as follows,

Step 4.1.1, for image to I_iAnd I_j, add up I respectively_iAnd I_jIn gray accumulation probability distribution graph CDF of overlapping region, K probit p is equally spaced chosen in main interval_k, image greyscale value corresponding for k ∈ [1, K], as sample point, is designated asWithK is default value；

Step 4.1.2, with I_iSample point and I_jSample point be that coordinate axes sets up plane right-angle coordinate ov_iv_j, obtain K from Scatterplot, simulates by Quadric Spline and obtains solid-line curve；

Step 4.1.3, calculates area between block curve section and datum line, is designated as A_s, then I_iAnd I_jGray difference value D_H(I_i,I_j)=A_s/(p_K-p₁)。

The present invention provides the color tone consistency correction system of a kind of sequence remote sensing image, including with lower module:

First module, for forwarding colored sequential images to L α β color space from RGB color；

Second module, transmits pretreatment for tone, including based on reference to image, utilizes the Wallis Filtering Formula of cum rights to shadow All pixels of picture carry out tone transmission respectively at passage L α β；

Three module, for luminance channel global gain compensate, including arrange with lower unit in luminance channel L to image greyscale Adjust,

Statistic unit, for adding up all adjacent images to the gray average μ (I in overlapped public territory_ij), wherein, I_ijRepresent With the lap of image j in image i；

Optimize unit, for the energy according to gain compensation, optimize the gray scale gain coefficient a of every image_i, i ∈ [1, N], wherein N is Image quantity；

Compensating unit, for the gain compensation factor optimum according to optimizing unit gained, does gain compensation to the luminance channel of every image It is as follows,

${\hat{I}}_i\left(p\right)=a_i{I}_i\left(p\right)$

Wherein, I_i(p) andRepresent arbitrary pixel p gray value before and after gain compensation on image i respectively；

4th module, for the Histogram Mapping of Color Channel, including arranging with lower unit at Color Channel α and β separately Image greyscale is adjusted,

Initial cell, is used for building undirected weighted graph, including individual image is considered as a node, is saved by the image with overlapping region Being connected with limit between point, limit power is set to this image to gray difference value D in overlapping region_H；

Judging unit, is used for setting threshold value T_D, work as D_H(I_i,I_j)≤T_D, then I_iAnd I_jLimit between image node is the consistent limit of tone, Otherwise, for the non-uniform limit of tone；

Search unit, for searching for the reference image set of optimum, is included in undirected weighted graph search and has the maximum on the consistent limit of tone Connected subgraph, the reference image set mapped as color channel histograms using the image node set of this subgraph；

Reference unit, for determining the directly reference image of every image, including by between the reference image node in undirected weighted graph Weights are set to 0, appoint and take one of them node as root node；Each image node is to root node to utilize shortest path first to determine Shortest path, obtain binary tree, the father node of each image node be exactly its carry out grey level histogram mapping directly with reference to shadow Picture；

Traversal Unit, for determining the grey scale mapping order of image sequence, including to the binary tree obtained in reference unit, from root node By the mode binary tree traversal of breadth First or depth-first, the image node sequence of actual access is just by gray scale between picture pair The execution sequence mapped；

Adjustment unit, for according to the execution sequence searched out, the Histogram Mapping relation between its reference image set up by every image, And adjust the gray scale of each pixel in image；

5th module, for all images are gone to RGB color from L α β color space, exports and preserves result.

And, adjacent image between the calculating process of gray difference value as follows,

For image to I_iAnd I_j, add up I respectively_iAnd I_jIn gray accumulation probability distribution graph CDF of overlapping region, at main region Between equally spaced choose K probit p_k, image greyscale value corresponding for k ∈ [1, K], as sample point, is designated asWithK is default value；

With I_iSample point and I_jSample point be that coordinate axes sets up plane right-angle coordinate ov_iv_j, obtain K discrete point, with two Secondary B-spline simulates and obtains solid-line curve；

Calculate area between block curve section and datum line, be designated as A_s, then I_iAnd I_jGray difference value D_H(I_i,I_j)=A_s/(p_K-p₁)。

The present invention utilizes the tone corresponding relation of overlapping region between adjacent image, and proposing one can effectively suppress or eliminate between image The brightness existed and the technical scheme of color distortion, the advantage of the technical program is: (1) a whole set of handling process is all at passage Between independence higher color space carry out, it is possible to largely reduce the image fault brought due to Inter-channel Correlation；(2) Brightness and color are separately processed, for respective characteristic, takes different processing schemes, can effectively promote the whole of sequence image Body color tone consistency.(3) the technical program is for color distortion problem present in sequential images, takes the maximum consistent figure of search As the mode of subset chooses reference color automatically, it is to avoid the problem of artificial selection standard reference color.

Accompanying drawing explanation

Fig. 1 is the process chart of the color tone consistency bearing calibration of the embodiment of the present invention carried sequence remote sensing image；

Fig. 2 is CDF curve equiprobability spacing sample point schematic diagram in the embodiment of the present invention, and transverse axis V represents that the gray scale of pixel takes Value, longitudinal axis P represents less than the pixel of certain gray value probability of occurrence in image；

Fig. 3 is that in the embodiment of the present invention, between image, Histogram distance describes schematic diagram, horizontally and vertically represents image i and image j The gray value that middle pixel is likely to occur；

Fig. 4 is the relevant classification limit non-directed graph of image sequence in the embodiment of the present invention, and node number represents the label of image, while represent Image between there is overlapping relation；

Fig. 5 be in the embodiment of the present invention image non-directed graph with reference to the image set minimum spanning tree as root node.

Detailed description of the invention

In order to make the purpose of the present invention, technical scheme and advantage clearer, below in conjunction with drawings and Examples, to this Bright it is further elaborated.Should be appreciated that specific embodiment described herein, only in order to explain the present invention, is not used to Limit the present invention.

The technical solution used in the present invention provides a kind of method that sequence remote sensing image is carried out light and color homogenization process.As it is shown in figure 1, By sequential images in the low L α β space of each Inter-channel Correlation, respectively luminance channel is carried out consistent adjustment with Color Channel, After transfer back to rgb space again and obtain final process result figure.Including following process step:

Step 1, forwards colored sequential images to L α β color space from RGB color.Transformation process is as follows:

Step 1.1, forwards the gray scale vector of pixel to LMS space, transition matrix T from rgb space₁:

$T_1=\left[\begin{array}{ccc}0.3811& 0.5783& -0.0787\\ -0.2298& 1.1834& 0.0464\\ 0.0000& 0.0000& 1.0000\end{array}\right]$

Step 1.2, forwards L α β space, transition matrix T to by the gray scale vector of pixel from LMS space₂:

$T_2=\left[\begin{array}{ccc}\frac{1}{\sqrt{3}}& 0& 0\\ 0& \frac{1}{\sqrt{6}}& 0\\ 0& 0& \frac{1}{\sqrt{2}}\end{array}\right]\left[\begin{array}{ccc}1& 1& 1\\ 1& 1& -2\\ 1& -1& 0\end{array}\right]$

Step 2, tone transmission pretreatment.A content can be pre-selected in or beyond sequential images and survey district situation is close and imaging Quality preferably remote sensing image as with reference to image, utilizes the Wallis Filtering Formula of the cum rights all pixels to image each in sequence Tone transmission is carried out respectively at L α β passage.Cum rights Wallis Filtering Formula is as follows:

$\hat{I}\left(p\right)=\frac{\mathrm{\σ}\left(I_r\right)}{\mathrm{\λ}\mathrm{\σ}\left(N_{n\×n}\right(p\left)\right)+(1-\mathrm{\λ})\mathrm{\σ}\left(I_r\right)}\left(I\right(p)-\mathrm{\μ}(N_{n\×n}\left(p\right)\left)\right)+\mathrm{\λ}\mathrm{\μ}\left(I_r\right)+(1-\mathrm{\λ})\mathrm{\μ}\left(N_{n\×n}\right(p\left)\right)$

In formula, μ () and σ () effect symbol represents the gray average to object pixel set and variance, Ν_n×n() effect symbol represents object pixel N × n neighborhood territory pixel set, when being embodied as, value n that those skilled in the art can be arranged voluntarily.I (p) andTable respectively Show the gray value after pixel p is before treatment in pending image, I_rRepresent all collection of pixels with reference to image.Weighting parameter λ ∈ [0,1] controls the intensity of this change action, and weights the biggest tone transmission effect is the most obvious, and vice versa.The present invention is at L α β The recommendation weights of passage are respectively 0.8,0.1,0.1.

Step 3, the global gain of luminance channel compensates.In L luminance channel, image greyscale is adjusted, first add up each adjacent The gray average of overlapping region between image, then use gain model that the brightness of image sequence carries out the concordance optimization of the overall situation.

The step 3 of embodiment implements and comprises the following steps:

Step 3.1, adds up all adjacent images to the gray average μ (I in overlapped public territory_ij), wherein, I_ijRepresent certain shadow As lap collection of pixels with adjacent image j in i.

Step 3.2, optimizes the gray scale gain coefficient a of every image_i, i ∈ [1, N], wherein N is image quantity.For certain image i and phase Adjacent image j, the energy equation of its gain compensation:

e_ij=(a_iμ(I_ij)-a_jμ(I_ji))²+α((a_i-1)²+(a_j-1)²)

Wherein, e_ijFor the energy of gain compensation, α is fixing weights, it is to avoid the trivial solution of equation, I_jiRepresent in certain image j Lap collection of pixels with adjacent image i.In view of all of image gray difference to lap, the energy side of the overall situation Formula:

$E={\mathrm{\Σ}}_{\underset{I_i\∩I_j\≠O}{i=1...N,j=1...N}}{e}_{ij}$

Above formula can pass through linear least square method rapid solving, i.e. obtains the gain compensation factor a of each image_i。

Step 3.3, according to the optimum gain penalty coefficient tried to achieve, does gain compensation to the luminance channel of every image: Wherein, I_i(p) andRepresent arbitrary pixel p gray value before and after gain compensation on image i respectively.

Step 4, the Histogram Mapping of Color Channel.Separately image greyscale is adjusted at α and β Color Channel, including According to the color distortion of overlapping region, search is optimum with reference to image set；Utilize shortest path first, determine the straight of every image Connect with reference to image and order；Set up and with reference to the Histogram Mapping relation between image, adjust image greyscale.

The step 4 of embodiment implements and is included in α and β passage and performs following steps respectively:

Step 4.1, builds undirected weighted graph.Every image is considered as a node (in Fig. 4, node number is the order label of image), Being connected having between the image node of overlapping region with limit, it is poor to the color (gray scale) in overlapping region that limit power is set to this image Different value D_H.Adjacent image between the calculating process of gray difference value following (with image to I_iAnd I_jAs a example by):

Step 4.1.1, adds up I respectively_iAnd I_jIn gray accumulation probability distribution graph CDF of overlapping region, in probit K probit p is equally spaced chosen between 0.5%～99.5%_k, image greyscale value corresponding for k ∈ [1, K], as sample point, is designated asWithAs shown in Figure 2.When being embodied as, those skilled in the art can preset the value of K voluntarily, Preferably value is 8.

Step 4.1.2, with I_iSample point and I_jSample point be that coordinate axes sets up plane right-angle coordinate ov_iv_j, then K is obtained Individual discrete point, simulates the solid-line curve in Fig. 3 by Quadric Spline.

Step 4.1.3, calculates area between block curve section and datum line (cross initial point and slope is the straight line of 1), is designated as A_s, Then I_iAnd I_jColor (gray scale) difference value D_H(I_i,I_j)=A_s/(p_K-p₁)。

Step 4.2, sets threshold value T_D, the limit of undirected weighted graph is divided into two classes: the consistent limit of tone and the non-uniform limit of tone.When D_H(I_i,I_j)≤T_D, then I_iAnd I_jLimit between image node is " the consistent limit of tone ", such as dotted line limit in Fig. 4；Otherwise, for " color Adjust non-uniform limit ", such as solid line limit in Fig. 4.When being embodied as, those skilled in the art can predetermined threshold value voluntarily, lead at α and β The recommendation threshold value in road is respectively 0.03 and 0.005.

Step 4.3, search is optimum with reference to image set.The maximal connected subgraphs with " the consistent limit of tone " is searched in undirected weighted graph, The reference image set mapped as color channel histograms using the image node set of this subgraph, such as the solid node in Fig. 4.

Step 4.4, determines the directly reference image of every image.Weights between reference image node in undirected weighted graph are set to 0, Appoint and take one of them node as root node.Shortest path first is utilized to determine each image node shortest path to root node, " binary tree " that obtain as shown in Figure 5.The father node of each image node is exactly that it carries out the direct reference of grey level histogram mapping Image.

Step 4.5, determines the grey scale mapping order of image sequence.To the binary tree obtained in step 4.4, press breadth First from root node Or the mode binary tree traversal of depth-first, the image node sequence of actual access is just by the execution of grey scale mapping between picture pair Sequentially.

Step 4.6, according to the execution sequence searched out, every image sets up the Histogram Mapping relation between its reference image, such as figure In 3, transverse and longitudinal coordinate is about the corresponding relation of solid-line curve, and adjusts the gray scale of each pixel in image according to this grey scale mapping relation.

All images are gone to RGB color from L α β color space by step 5, and conversion regime is the inverse operation of step 1, Export and preserve result, obtaining final result image sequence.

When being embodied as, method provided by the present invention can realize automatic operational process based on software engineering, it is possible to uses modularity side Formula realizes corresponding system.The embodiment of the present invention also provides for the color tone consistency correction system of a kind of sequence remote sensing image, including following Module:

First module, for forwarding colored sequential images to L α β color space from RGB color；

Second module, transmits pretreatment for tone, including based on reference to image, utilizes the Wallis Filtering Formula of cum rights to shadow All pixels of picture carry out tone transmission respectively at passage L α β；

Three module, for luminance channel global gain compensate, including arrange with lower unit in luminance channel L to image greyscale Adjust,

Statistic unit, for adding up all adjacent images to the gray average μ (I in overlapped public territory_ij), wherein, I_ijRepresent With the lap of image j in image i；

Optimize unit, for the energy according to gain compensation, optimize the gray scale gain coefficient a of every image_i, i ∈ [1, N], wherein N is Image quantity；

Compensating unit, for the gain compensation factor optimum according to optimizing unit gained, does gain compensation to the luminance channel of every image It is as follows,

${\hat{I}}_i\left(p\right)=a_i{I}_i\left(p\right)$

Wherein, I_i(p) andRepresent arbitrary pixel p gray value before and after gain compensation on image i respectively；

4th module, for the Histogram Mapping of Color Channel, including arranging with lower unit at Color Channel α and β separately Image greyscale is adjusted,

Initial cell, is used for building undirected weighted graph, including individual image is considered as a node, is saved by the image with overlapping region Being connected with limit between point, limit power is set to this image to gray difference value D in overlapping region_H；

Judging unit, is used for setting threshold value T_D, work as D_H(I_i,I_j)≤T_D, then I_iAnd I_jLimit between image node is the consistent limit of tone, Otherwise, for the non-uniform limit of tone；

Search unit, for searching for the reference image set of optimum, is included in undirected weighted graph search and has the maximum on the consistent limit of tone Connected subgraph, the reference image set mapped as color channel histograms using the image node set of this subgraph；

Reference unit, for determining the directly reference image of every image, including by between the reference image node in undirected weighted graph Weights are set to 0, appoint and take one of them node as root node；Each image node is to root node to utilize shortest path first to determine Shortest path, obtain binary tree, the father node of each image node be exactly its carry out grey level histogram mapping directly with reference to shadow Picture；

Traversal Unit, for determining the grey scale mapping order of image sequence, including to the binary tree obtained in reference unit, from root node By the mode binary tree traversal of breadth First or depth-first, the image node sequence of actual access is just by gray scale between picture pair The execution sequence mapped；

Adjustment unit, for according to the execution sequence searched out, the Histogram Mapping relation between its reference image set up by every image, And adjust the gray scale of each pixel in image；

5th module, for all images are gone to RGB color from L α β color space, exports and preserves result.

Each module implements and can be found in corresponding steps, and it will not go into details for the present invention.

Specific embodiment described herein is only to present invention spirit explanation for example.The skill of the technical field of the invention Described specific embodiment can be made various amendment or supplements or use similar mode to substitute by art personnel, but not The spirit of the present invention can be deviateed or surmount scope defined in appended claims.

Claims (4)

1. the color tone consistency bearing calibration of a sequence remote sensing image, it is characterised in that comprise the following steps:

Step 1, forwards colored sequential images to L α β color space from RGB color；

Step 2, tone transmission pretreatment, including based on reference to image, utilize the Wallis Filtering Formula institute to image of cum rights Pixel is had to carry out tone transmission respectively at passage L α β；

Step 3, the global gain of luminance channel compensates, adjusts image greyscale in luminance channel L including performing following sub-step It is whole,

Step 3.1, adds up all adjacent images to the gray average μ (I in overlapped public territory_ij), wherein, I_ijRepresent image i In with the lap of image j；

Step 3.2, according to the energy of gain compensation, optimizes the gray scale gain coefficient a of every image_i,i∈[1,N]_,Its_InN is image number Amount；

Step 3.3, according to the gain compensation factor that step 3.2 gained is optimum, does gain compensation to the luminance channel of every image as follows,

${\hat{I}}_i\left(p\right)=a_i{I}_i\left(p\right)$

Wherein, I_i(p) andRepresent arbitrary pixel p gray value before and after gain compensation on image i respectively；

Step 4, the Histogram Mapping of Color Channel, including performing following sub-step at Color Channel α and β separately to shadow As gray scale adjusts,

Step 4.1, builds undirected weighted graph, including individual image is considered as a node, will have between the image node of overlapping region Being connected with limit, limit power is set to this image to gray difference value D in overlapping region_H；

Step 4.2, sets threshold value T_D, work as D_H(I_i,I_j)≤T_D, then I_iAnd I_jLimit between image node is the consistent limit of tone, otherwise, For the non-uniform limit of tone；

Step 4.3, the reference image set that search is optimum, it is included in undirected weighted graph search and there is the largest connected son on the consistent limit of tone Figure, the reference image set mapped as color channel histograms using the image node set of this subgraph；

Step 4.4, determines the directly reference image of every image, sets including by the weights between the reference image node in undirected weighted graph It is 0, appoints and take one of them node as root node；Shortest path first is utilized to determine the shortest to root node of each image node Path, the father node of each image node is exactly that it carries out the directly reference image of grey level histogram mapping；

Step 4.5, determines the grey scale mapping order of image sequence, including to the binary tree obtained in step 4.4, presses range from root node Preferential or the mode binary tree traversal of depth-first, the image node sequence of actual access be just by as pair between grey scale mapping Execution sequence；

Step 4.6, according to the execution sequence searched out, every image is set up the Histogram Mapping relation between its reference image, and is adjusted The gray scale of each pixel in whole image；

All images are gone to RGB color from L α β color space, export and preserve result by step 5.

The color tone consistency bearing calibration of sequence remote sensing image the most according to claim 1, it is characterised in that: adjacent image between The calculating process of gray difference value is as follows,

Step 4.1.1, for image to I_iAnd I_j, add up I respectively_iAnd I_jIn gray accumulation probability distribution graph CDF of overlapping region, K probit p is equally spaced chosen in main interval_k, image greyscale value corresponding for k ∈ [1, K], as sample point, is designated asWithK is default value；

Step 4.1.2, with I_iSample point and I_jSample point be that coordinate axes sets up plane right-angle coordinate ov_iv_j, obtain K from Scatterplot, simulates by Quadric Spline and obtains solid-line curve；

Step 4.1.3, calculates area between block curve section and datum line, is designated as A_s, then I_iAnd I_jGray difference value D_H(I_i,I_j)=A_s/(p_K-p₁)。

3. the color tone consistency correction system of a sequence remote sensing image, it is characterised in that include with lower module:

First module, for forwarding colored sequential images to L α β color space from RGB color；

Second module, transmits pretreatment for tone, including based on reference to image, utilizes the Wallis Filtering Formula of cum rights to shadow All pixels of picture carry out tone transmission respectively at passage L α β；

Three module, for luminance channel global gain compensate, including arrange with lower unit in luminance channel L to image greyscale Adjust,

Statistic unit, for adding up all adjacent images to the gray average μ (I in overlapped public territory_ij), wherein, I_ijRepresent With the lap of image j in image i；

Optimize unit, for the energy according to gain compensation, optimize the gray scale gain coefficient a of every image_i, i ∈ [1, N], wherein N is Image quantity；

Compensating unit, for the gain compensation factor optimum according to optimizing unit gained, does gain compensation to the luminance channel of every image It is as follows,

${\hat{I}}_i\left(p\right)=a_i{I}_i\left(p\right)$

Wherein, I_i(p) andRepresent arbitrary pixel p gray value before and after gain compensation on image i respectively；

4th module, for the Histogram Mapping of Color Channel, including arranging with lower unit at Color Channel α and β separately Image greyscale is adjusted,

Initial cell, is used for building undirected weighted graph, including individual image is considered as a node, is saved by the image with overlapping region Being connected with limit between point, limit power is set to this image to gray difference value D in overlapping region_H；

Judging unit, is used for setting threshold value T_D, work as D_H(I_i,I_j)≤T_D, then I_iAnd I_jLimit between image node is the consistent limit of tone, Otherwise, for the non-uniform limit of tone；

Search unit, for searching for the reference image set of optimum, is included in undirected weighted graph search and has the maximum on the consistent limit of tone Connected subgraph, the reference image set mapped as color channel histograms using the image node set of this subgraph；

Reference unit, for determining the directly reference image of every image, including by between the reference image node in undirected weighted graph Weights are set to 0, appoint and take one of them node as root node；Each image node is to root node to utilize shortest path first to determine Shortest path, obtain binary tree, the father node of each image node be exactly its carry out grey level histogram mapping directly with reference to shadow Picture；

Traversal Unit, for determining the grey scale mapping order of image sequence, including to the binary tree obtained in reference unit, from root node By the mode binary tree traversal of breadth First or depth-first, the image node sequence of actual access is just by gray scale between picture pair The execution sequence mapped；

Adjustment unit, for according to the execution sequence searched out, the Histogram Mapping relation between its reference image set up by every image, And adjust the gray scale of each pixel in image；

5th module, for all images are gone to RGB color from L α β color space, exports and preserves result.

The most according to claim 3 sequence remote sensing image color tone consistency correction system, it is characterised in that: adjacent image between The calculating process of gray difference value is as follows,

For image to I_iAnd I_j, add up I respectively_iAnd I_jIn gray accumulation probability distribution graph CDF of overlapping region, at main region Between equally spaced choose K probit p_k,Image greyscale value corresponding for k ∈ [1, K], as sample point, is designated asWithK is default value；

With I_iSample point and I_jSample point be that coordinate axes sets up plane right-angle coordinate ov_iv_j, obtain K discrete point, with two Secondary B-spline simulates and obtains solid-line curve；

Calculate area between block curve section and datum line, be designated as A_s, then I_iAnd I_jGray difference value D_H(I_i,I_j)=A_s/(p_K-p₁)。