CN-116342425-B - Area array satellite image relative radiation correction method considering regional response difference

CN116342425BCN 116342425 BCN116342425 BCN 116342425BCN-116342425-B

Abstract

The invention discloses an area array satellite image relative radiation correction method considering regional response difference. According to the method, firstly, coarse features of an image are removed based on NSCT transformation, then fine features of the image are removed through RCV, the image mainly containing noise is obtained, a priori model considering regional response difference is constructed to estimate a relative radiation correction coefficient, and relative radiation correction is carried out on other images obtained by the same sensor under the constraint of a corresponding cloud mask, so that imaging quality degradation of an area array satellite image due to inconsistent probe response is eliminated, and the radiation quality of the area array satellite image is improved. The invention can effectively reduce the number of the area array satellite images required by estimating the relative radiation correction coefficient by removing the coarse features and the fine features, and breaks through the limitation that the area array satellite images used for estimating the relative radiation correction coefficient and carrying out the relative radiation correction must be cloud-free images because the influence of the regional response difference is considered.

Inventors

PAN JUN
Wan Tianzhen
WANG MI

Assignees

武汉大学

Dates

Publication Date: 20260512
Application Date: 20230330

Claims (9)

1. An area array satellite image relative radiation correction method considering area response difference is characterized by comprising the following steps: Step 1, removing coarse features of an image by utilizing non-downsampled contourlet transform NSCT; step 1.1, decomposing an original image into a low-frequency sub-band and a plurality of high-frequency sub-bands by using NSCT transformation; After the original image is decomposed by NSCT transformation, a low-frequency sub-band and a plurality of high-frequency sub-bands are obtained, coarse features mainly exist in the low-frequency sub-bands, edges, contour information and noise of the image are contained in the high-frequency sub-bands, the image can be regarded as being composed of two parts of feature information and noise, wherein the feature information is divided into coarse features and fine features, the noise is divided into random noise and system noise, and the composition of an image is expressed by the following formula in consideration of the fact that the system noise is a multiplicative noise: (1) In the formula, Representing the original image in Is used for the gray-scale value of (c), Is a coarse characteristic image The gray value at which the color is to be changed, Representing fine feature images in The gray value at which the color is to be changed, Is a random noise which is a random noise, Representing system noise image in Gray value of the position, system noise acts on the image without noise according to multiplicative rule; Step 1.2, updating low-frequency sub-band coefficients by bilateral filtering to obtain coarse features in the low-frequency sub-band, updating the low-frequency sub-band coefficients through a layering threshold value, judging whether the high-frequency sub-band coefficients correspond to edges and contours of an image or not, and filtering out edge information and image contour information as noise so as to retain the coarse features in the high-frequency sub-band; step 1.3, performing NSCT inverse transformation based on the updated low-frequency sub-band coefficients and the updated high-frequency sub-band coefficients, reconstructing coarse features of an original image, and removing the coarse features of the image; Step 2, removing fine features of the image based on the regional covariance matrix; step 2.1, performing visual feature coding on the image with the rough features removed; Step 2.2, measuring the similarity between center pixels of the image block by utilizing the regional covariance matrix and combining with the visual feature codes; Step 2.3, extracting and removing fine features to obtain an image only containing system noise and random noise; Step 3, constructing and using a priori model taking the regional response difference into consideration to acquire a relative radiation correction coefficient, and performing relative radiation correction processing on other images acquired by the same sensor; Step 3.1, constructing a priori model considering the regional response difference; Considering the influence of abnormal values generated by a cloud region on the estimation of the relative radiation correction coefficient, constructing a prior model considering the difference of the regional response as follows: (11) In the formula, And Respectively represent the square of the vector L2 norm and the L1 norm, Representing the relative radiation correction factor, whose value is the inverse of the system noise, The size of the image is indicated and, In the form of a diagonal matrix, Representing the number of images used to calculate the relative radiation correction factor, The regularization parameters are represented by a set of values, Representing the primary correction of the original image without removing the coarse features and the fine features, and the coefficients 、 Is defined as follows: In the formula, Is an identity matrix of the unit cell, Representing the size of the image; In the formula, Representing noisy images Is the inverse of the number of (a), And The height and width of the image are shown separately, The size of the image is indicated and, Obtained by the same sensor Individual vectors Combining; Step 3.2, optimizing a priori model taking the regional response difference into consideration, and eliminating random noise; And 3.3, carrying out relative radiation correction on other images acquired by the same sensor according to the estimated relative radiation correction coefficient.
2. The method of claim 1, wherein the step 1.3 of removing the rough features is performed as follows; (2) In the formula, Representing images with coarse features removed The gray value at which the color is to be changed, Representing the original image in Is used for the gray-scale value of (c), Is a coarse characteristic image The gray value at which the color is to be changed, Representing fine feature images in The gray value at which the color is to be changed, Is a random noise which is a random noise, Representing system noise image in Gray values at that point.
3. The method of claim 1, wherein the step 2.1 of dividing the image with coarse features removed into images of the same size and overlapping each other is Then, each image block is subjected to visual feature coding, and the coding formula is as follows: (3) In the formula, Expressed in terms of position Is of the size of the center The visual characteristics of the image block, , , Each row of the image block represents a respective pixel within the image block, each column corresponds to 7 visual features of the pixel, Representing the vectorized intensity values of the image block at (X, Y), And Respectively by discrete filters Calculated as The first derivative in the horizontal and vertical directions, And Respectively represent the utilization of discrete filters Calculated as The second derivative in the horizontal and vertical directions, And Respectively representing coordinates of pixels within the image block.
4. The method for correcting the relative radiation of an area-response-difference-considered area array satellite image as claimed in claim 3, wherein the similarity between center pixels of the image block in step 2.2 is calculated as follows: (4) In the formula, Representing the position And A similarity between two pixels is present, Expressed in terms of position Is of the size of the center The average value of the visual characteristics of all pixels within an image block, Expressed in terms of position Is of the size of the center The average value of the visual characteristics of all pixels within an image block, The covariance descriptor extracted from the regional covariance matrix is represented by the following specific calculation method: (5) (6) (7) In the formula, 、 Respectively by position 、 Is of the size of the center The visual characteristics of the image block, 、 Representing image blocks respectively 、 The average value of the visual characteristics of all pixels within.
5. The method for correcting the relative radiation of the area-response-difference-considered area array satellite image according to claim 2, wherein the fine feature image in step 2.3 is specifically calculated as follows: (8) In the formula, Representing fine feature images in The gray value at which the color is to be changed, Representing images with coarse features removed Gray value at, pixel In pixels In the neighborhood with the center and the radius r, the weight is as follows Representing two pixels And The distance between them is defined as follows: (9) In the formula, Representing the position And A similarity between two pixels is present, Representing the smoothing parameters.
6. The method of claim 5, wherein the noise image without coarse and fine features in step 2.3 is calculated by the following formula: (10) In the formula, Representing an image containing both systematic noise and random noise, Representing images with coarse features removed The gray value at which the color is to be changed, Representing the original image in Is used for the gray-scale value of (c), Is a coarse characteristic image The gray value at which the color is to be changed, Representing fine feature images in The gray value at which the color is to be changed, Is a random noise which is a random noise, Representing system noise image in Gray values at that point.
7. The method of claim 1, wherein the area response difference is taken into account by an area array satellite image relative radiation correction method, comprising: in step 3.1 And Is defined as follows: (12) (13) In the formula, Representing the first order gradient of the preliminary corrected image, Representation of In position Is used for the display of the display panel, Representation of In position X represents the pixel position; In the formula, From the following components Diagonal matrix of each Is combined with the components of the composite material, The size of the image is indicated and, Is defined as follows: In the formula, Representing the raw image acquired by the same sensor, Representing the size of the image.
8. The method of claim 7, wherein the step 3.2 is performed by introducing an auxiliary variable 、z、 And define , The prior model taking into account the regional response differences can be converted into the following form: (18) the minimization of equation (18) may be further translated into the iterative step of: (19) (20) (21) (22) (23) In the formula, The number of iterations is indicated and, And Representing the augmented lagrangian multiplier, And Representing the augmented lagrangian coefficient.
9. The method of claim 8, wherein the area response difference is taken into account by an area array satellite image relative radiation correction method, comprising: in step 3.3, the relative radiation correction coefficient is obtained Then, the original image is corrected by Obtaining a preliminary relative radiation correction image Then, replacing gray values of the cloud area with uncorrected original pixel values to obtain a final relative radiation correction result; (24) in which the symbols are The point-of-view is indicated, Representing the relative radiation correction factor.

Description

Area array satellite image relative radiation correction method considering regional response difference Technical Field The invention belongs to the technical field of relative radiation correction of remote sensing images, and particularly relates to an area array satellite image relative radiation correction method considering regional response differences. Background In the imaging process of the area array satellite image, the non-uniform radiation distortion caused by inconsistent response of the detector in space, time instability and circuit noise not only reduces the visual quality of the image, but also can influence the analysis and subsequent processing of the image. The research on the relative radiation correction of the area array satellite eliminates the imaging quality degradation of the area array satellite image caused by non-uniform radiation distortion, and has important significance for improving the radiation quality of the area array satellite image. For the relative radiation correction of the area array satellite image, most correction methods are based on laboratory radiation calibration at present, and the response relationship between the gray value of the area array satellite image and the irradiance of an outlet of an integrating sphere is established by utilizing different radiance generated by the integrating sphere. Common methods are single point correction, two point correction and multi point correction. Since the single point correction algorithm compensates only for the offset of each pixel, the correction effect may be deteriorated when the radiance of the target deviates from the calibration point. The effect of two-point correction is limited by the dynamic range, and when the response inconsistency between pixels is large, the correction may deviate. The multi-point correction algorithm is essentially an extension of two-point correction, and approximates a nonlinear response characteristic curve by using multi-point correction, so that the accuracy is high, but the implementation is complex and difficult. In order to solve the above problems, a method of correcting relative radiation using a plurality of images has been proposed, which removes characteristics of images by overlapping a large number of images while preserving and enhancing system noise as commonality of images, but which requires a huge number of input images to support, and causes coarse characteristics, fine characteristics, and random noise of images to gradually disappear during overlapping of a large number of images. The method requires that the input image is a cloud-free image or has little cloud quantity, because the gray value of the cloud area is more continuous, the texture change is not obvious, the original signal to noise ratio is high, and the response difference between the original signal to noise ratio and the image of the cloud-free area is larger, so that the effect of final relative radiation correction can be influenced. In an area array satellite image, non-uniform radiation distortion can exist in the form of point noise in multiple images. The same point noise should be present for all images acquired by the same sensor, i.e. the point noise is a system noise, so that the problem of area array relative radiation correction can be converted into a system noise for removing multiple images. Disclosure of Invention Aiming at the defects of the prior art, the invention provides an area array satellite image relative radiation correction method considering the difference of area response. The method can effectively reduce the number of the area array satellite images required by estimating the relative radiation correction coefficient by removing the coarse features and the fine features, and simultaneously breaks through the limitation that the area array satellite images used for estimating the relative radiation correction coefficient and carrying out the relative radiation correction must be cloud-free images due to the consideration of the influence of regional response difference. In order to achieve the above purpose, the technical scheme provided by the invention is an area array satellite image relative radiation correction method considering the difference of area response, comprising the following steps: Step 1, removing coarse features of an image by utilizing non-downsampled contourlet transform NSCT; step 1.1, decomposing an original image into a low-frequency sub-band and a plurality of high-frequency sub-bands by using NSCT transformation; Step 1.2, updating low-frequency sub-band coefficients by bilateral filtering to obtain coarse features in the low-frequency sub-band, updating the low-frequency sub-band coefficients through a layering threshold value, judging whether the high-frequency sub-band coefficients correspond to edges and contours of an image or not, and filtering out edge information and image contour information as noise so as to retain the coarse