CN103679653B - A kind of satellite image veiling glare eliminates system and method - Google Patents

Abstract

The invention discloses a satellite image stray light elimination system and method, including a data acquisition module, an image registration module and a stray light suppression module; wherein the data acquisition module includes an image grayscale acquisition unit and an image parameter acquisition unit, and an image registration module It includes a coarse image registration unit and a fine image registration unit, and the stray light suppression module includes a proportional coefficient calculation unit, a difference image calculation unit and an image filtering unit. The present invention first registers the high-resolution camera image and the wide-format camera image that are simultaneously imaged on-orbit, and establishes the correspondence between the gray-scale value of the high-resolution camera image and the gray-scale value of the wide-format camera image according to the spectral correspondence between the images Relational model to eliminate the influence of stray light in the image. Using this method to eliminate stray light in the image is simple and practical, it can preserve the real information of the image to a large extent, and the effect of removing stray light is good.

Description

A satellite image stray light elimination system and method

technical field

The invention relates to a satellite image stray light elimination system and method, belonging to the satellite application field.

Background technique

Stray light brightens the background of satellite images, reduces the image contrast, modulation transfer function and signal-to-noise ratio, has a serious impact on the imaging quality of the optical system, and brings difficulties to the interpretation and application of satellite images. It is necessary to analyze and eliminate the stray light of satellite images, so as to ensure the imaging quality of satellite optical payloads.

Much of the effort in stray light suppression is achieved through optical system design. The most fundamental way to eliminate stray light is to study the source of stray light and then analyze the geometrical optics policy. The calculation methods of stray light mainly include Monte Carlo method, area method, ray tracing method and so on. Among them, the Monte Carlo method is the only relatively mature method in the field of stray light analysis, but it needs to obtain the physical model of each process of light propagation. For design omissions or the influence of stray light that cannot be eliminated due to the complex space environment, it can only be eliminated by the back-end image processing method. Yuan Yukai et al. used the satellite imaging data at a specific time to separate the real signal and the stray light signal, generated the stray light template at each time, and suppressed the stray light on the visible light image of the FY-2 radiometer through the image processing method of deconvolution. However, this method is only suitable for geosynchronous meteorological satellites, and for medium and high-resolution earth observation satellite data, it cannot separate stray light signals based on imaging data at a specific time.

At present, there are few methods for directly suppressing stray light based on the image domain, and there is no method for suppressing stray light in the image domain of medium and high-resolution earth observation satellite images in practical applications. The influence of stray light that cannot be removed by radiation correction in medium and high-resolution remote sensing images cannot be solved.

Contents of the invention

The technical problem to be solved by the present invention is to provide a satellite image stray light elimination system and method to overcome the deficiencies in the prior art, so as to eliminate image stray light and preserve image real information to a large extent.

Technical solution of the present invention: a satellite image stray light elimination system, including a data acquisition module, an image registration module and a stray light suppression module; wherein the data acquisition module includes an image gray scale acquisition unit and an image parameter acquisition unit, and the image registration module The module includes an image coarse registration unit and an image fine registration unit, and the stray light suppression module includes a proportional coefficient calculation unit, a difference image calculation unit and an image filtering unit;

The data acquisition module obtains high-resolution camera and wide-format camera images and image-related parameters in the same area from the image database, and defines one of the images as an image to be processed, and another image as a grayscale reference image; image registration The module takes the image to be processed as a reference image, and registers the image to be processed and the gray scale reference image; the stray light suppression module removes the stray light in the image to be processed;

The image grayscale acquisition unit acquires the grayscale information of the image to be processed and the grayscale reference image from the image database; the image parameter acquisition unit acquires the parameter information of the image to be processed and the grayscale reference image, and the image to be processed and the grayscale reference image The grayscale information and parameter information are sent to the rough image registration unit in the image registration module;

The image coarse registration unit obtains the grayscale information and parameter information of the image to be processed and the grayscale reference image provided by the data acquisition module, eliminates the resolution difference between the grayscale reference image and the image to be processed, and outputs the image to be processed and the grayscale Reference image; the image fine registration unit selects control points in the image to be processed, selects the point with the same name as the aforementioned control point in the grayscale reference image, uses an affine transformation model, resamples the grayscale reference image, and obtains an image similar to the image to be processed The finely-registered grayscale reference image, outputting the grayscale reference image and the image to be processed to the proportional coefficient calculation unit of the stray light suppression module;

The proportional coefficient calculation unit calculates the ratio of the pixel gray value of the corresponding position of the image to be processed and the grayscale reference image to obtain a proportional coefficient matrix, and optimizes the proportional coefficient matrix, and outputs the optimized proportional coefficient matrix to the difference image calculation unit; The image calculation unit obtains the difference image through calculation, and sends the difference image to the image filter unit; the image filter unit performs filtering processing on the difference image to obtain the filtered difference image; subtracts the filtered difference image from the image to be processed , get the stray light suppression result image and output the image to the database.

The method to eliminate the resolution difference between the grayscale reference image and the image to be processed is to divide the resolution of the grayscale reference image in the parameter information of the grayscale reference image by the resolution of the image to be processed to obtain the resolution between the two images Resolution difference, based on the image to be processed, resampling the grayscale reference pattern to make the grayscale reference image consistent with the resolution of the image to be processed.

The calculation method of the difference image calculation unit to obtain the difference image is as follows: firstly, the gray value of each pixel of the gray reference image is multiplied by the scale coefficient corresponding to each pixel position in the optimized scale coefficient matrix to obtain the reconstructed to-be-processed The ideal image of the image; the ideal image of the image to be processed is subtracted from the image to be processed to obtain a difference image.

A method for eliminating stray light from a satellite image, comprising the following steps:

Step 1, the data acquisition module acquires the high-resolution camera and wide-format camera images and image related parameters in the same area, and defines one of the images as the image to be processed, and the other as the grayscale reference image; the data acquisition module includes image grayscale A degree acquisition unit and an image parameter acquisition unit, the image grayscale acquisition unit reads the images I _H and I _W of the high-resolution camera and the wide-format camera, and the image parameter acquisition unit reads the imaging line-time parameters and parameters of the high-resolution camera and the wide-format camera Two camera image resolution parameters S _H , S _W define the high-resolution camera image as the image to be processed, and the wide-format camera image as the grayscale reference image; according to the imaging time, intercept the grayscale reference image I _W and the image to be processed I _H The corresponding area is defined as I _WR , and I _WR must include I _H ;

Step 2. According to the image parameter information, the image registration module uses the manual point selection method to use the image to be processed as a reference image to register the image to be processed and the grayscale reference image; the image registration module includes an image coarse registration unit and an image In the fine registration unit, in the image coarse registration unit, the proportional relationship P _WS between the two images is calculated by S _W / _SH , and the I _WR is resampled by the bicubic sampling method to obtain a new grayscale reference image I _HR , the resolution of which is the same as I _H is consistent; the image fine registration unit manually selects r rows and c columns of r×c control points uniformly on I _H and I _HR , and uses I _H as a reference to register I _H and I _HR using affine transformation , to obtain the grayscale reference image I _HR ' that is completely aligned with the spatial position of I _H .

Step 3, according to the spectral corresponding relationship between the grayscale reference image and the image to be processed, the stray light suppression module suppresses the stray light in the image to be processed to obtain the stray light suppression result image; the stray light suppression module includes a proportional coefficient calculation unit, a difference Value image calculation unit and image microwave unit, in the proportional coefficient calculation unit, add 0.0001 to the gray value of each pixel in I _HR ', divide each pixel with the same spatial position in I _H and I _HR ', and obtain the proportional coefficient matrix Coff _HW , elements greater than threshold T _CoffH and less than threshold T _CoffL in the matrix, make it equal to 1; calculate the image average Mean _H and Mean _HR of I _H and I _HR ' in the scale factor calculation unit, scale reference value Coff _C = Mean _H /Mean _HR , elements greater than Coff _C in Coff _HW , make its value equal to Coff _C ; in the difference image calculation unit, the gray value of each pixel in the gray reference image I _HR ' is multiplied by the proportional coefficient matrix Coff The scale coefficient corresponding to its spatial position in _HW , calculate the difference image Diff=I _H -I _HR ', the element whose gray value absolute value is greater than the threshold T _DiffH and less than the threshold T _DiffL in Diff is equal to 0; in the image The filter unit selects a square mean value filter template whose side length is L, and filters the Diff mean value; subtracts I _H from the gray value of the pixel corresponding to the space in the Diff to obtain a stray light suppression result image.

The beneficial effects of the present invention compared with prior art are:

The present invention first registers the high-resolution camera image and the wide-format camera image that are simultaneously imaged on-orbit, and establishes the correspondence between the gray-scale value of the high-resolution camera image and the gray-scale value of the wide-format camera image according to the spectral correspondence between the images Relational model to eliminate the influence of stray light in the image. Using this method to eliminate stray light in the image is simple and practical, it can preserve the real information of the image to a large extent, and the effect of removing stray light is good.

Description of drawings

Fig. 1 is a schematic diagram of the system of the present invention;

Fig. 2 is a schematic diagram of the method of the present invention.

detailed description

Below in conjunction with accompanying drawing 1,2 the present invention is described in further detail.

A satellite image stray light elimination system, including a data acquisition module, an image registration module and a stray light suppression module; wherein the data acquisition module includes an image grayscale acquisition unit and an image parameter acquisition unit, and the image registration module includes an image coarse registration unit and an image fine registration unit, the stray light suppression module includes a proportional coefficient calculation unit, a difference image calculation unit and an image filtering unit;

The data acquisition module obtains high-resolution camera and wide-format camera images and image-related parameters in the same area from the image database, and defines one of the images as an image to be processed, and another image as a grayscale reference image; image registration The module takes the image to be processed as a reference image, and registers the image to be processed and the gray scale reference image; the stray light suppression module removes the stray light in the image to be processed;

The image grayscale acquisition unit acquires the grayscale information of the image to be processed and the grayscale reference image from the image database; the image parameter acquisition unit acquires the parameter information of the image to be processed and the grayscale reference image, and the image to be processed and the grayscale reference image The grayscale information and parameter information are sent to the rough image registration unit in the image registration module;

The image coarse registration unit obtains the grayscale information and parameter information of the image to be processed and the grayscale reference image provided by the data acquisition module, eliminates the resolution difference between the grayscale reference image and the image to be processed, and outputs the image to be processed and the grayscale Reference image; the image fine registration unit selects control points in the image to be processed, selects the point with the same name as the aforementioned control point in the grayscale reference image, uses an affine transformation model, resamples the grayscale reference image, and obtains an image similar to the image to be processed The finely-registered grayscale reference image, outputting the grayscale reference image and the image to be processed to the proportional coefficient calculation unit of the stray light suppression module;

The proportional coefficient calculation unit calculates the ratio of the pixel gray value of the corresponding position of the image to be processed and the grayscale reference image to obtain a proportional coefficient matrix, and optimizes the proportional coefficient matrix, and outputs the optimized proportional coefficient matrix to the difference image calculation unit; The image calculation unit obtains the difference image through calculation, and sends the difference image to the image filter unit; the image filter unit performs filtering processing on the difference image to obtain the filtered difference image; subtracts the filtered difference image from the image to be processed , get the stray light suppression result image and output the image to the database.

The method to eliminate the resolution difference between the grayscale reference image and the image to be processed is to divide the resolution of the grayscale reference image in the parameter information of the grayscale reference image by the resolution of the image to be processed to obtain the resolution between the two images Resolution difference, based on the image to be processed, resampling the grayscale reference pattern to make the grayscale reference image consistent with the resolution of the image to be processed.

The calculation method of the difference image calculation unit to obtain the difference image is as follows: firstly, the gray value of each pixel of the gray reference image is multiplied by the scale coefficient corresponding to each pixel position in the optimized scale coefficient matrix to obtain the reconstructed to-be-processed The ideal image of the image; the ideal image of the image to be processed is subtracted from the image to be processed to obtain a difference image.

A method for eliminating stray light from a satellite image, comprising the following steps:

Step 1, the data acquisition module acquires the high-resolution camera and wide-format camera images and image related parameters in the same area, and defines one of the images as the image to be processed, and the other as the grayscale reference image; the data acquisition module includes image grayscale A degree acquisition unit and an image parameter acquisition unit, the image grayscale acquisition unit reads the images I _H and I _W of the high-resolution camera and the wide-format camera, and the image parameter acquisition unit reads the imaging line-time parameters and parameters of the high-resolution camera and the wide-format camera Two camera image resolution parameters S _H , S _W define the high-resolution camera image as the image to be processed, and the wide-format camera image as the grayscale reference image; according to the imaging time, intercept the grayscale reference image I _W and the image to be processed I _H The corresponding area is defined as I _WR , and I _WR must include I _H ;

Step 2. According to the image parameter information, the image registration module uses the manual point selection method to use the image to be processed as a reference image to register the image to be processed and the grayscale reference image; the image registration module includes an image coarse registration unit and an image In the fine registration unit, in the image coarse registration unit, the proportional relationship P _WS between the two images is calculated by S _W / _SH , and the I _WR is resampled by the bicubic sampling method to obtain a new grayscale reference image I _HR , the resolution of which is the same as I _H is consistent; the image fine registration unit manually selects r rows and c columns of r×c control points uniformly on I _H and I _HR , and uses I _H as a reference to register I _H and I _HR using affine transformation , to obtain the grayscale reference image I _HR ' that is completely aligned with the spatial position of I _H .

Step 3, according to the spectral corresponding relationship between the grayscale reference image and the image to be processed, the stray light suppression module suppresses the stray light in the image to be processed to obtain the stray light suppression result image; the stray light suppression module includes a proportional coefficient calculation unit, a difference Value image calculation unit and image microwave unit, in the proportional coefficient calculation unit, add 0.0001 to the gray value of each pixel in I _HR ', divide each pixel with the same spatial position in I _H and I _HR ', and obtain the proportional coefficient matrix Coff _HW , elements greater than threshold T _CoffH and less than threshold T _CoffL in the matrix, make it equal to 1; calculate the image average Mean _H and Mean _HR of I _H and I _HR ' in the scale factor calculation unit, scale reference value Coff _C = Mean _H /Mean _HR , elements greater than Coff _C in Coff _HW , make its value equal to Coff _C ; in the difference image calculation unit, the gray value of each pixel in the gray reference image I _HR ' is multiplied by the proportional coefficient matrix Coff The scale coefficient corresponding to its spatial position in _HW , calculate the difference image Diff=I _H -I _HR ', the element whose gray value absolute value is greater than the threshold T _DiffH and less than the threshold T _DiffL in Diff is equal to 0; in the image The filter unit selects a square mean value filter template whose side length is L, and filters the Diff mean value; subtracts I _H from the gray value of the pixel corresponding to the space in the Diff to obtain a stray light suppression result image.

Specifically:

1. Obtain image and image parameters

1.1 The high-resolution camera and wide-format camera images I _H and I _W to be processed are read by the image acquisition unit, wherein the size of I _H is (2865 pixels × 3036 pixels), and the size of I _W is (2287 pixels × 12000 pixels). The parameter acquisition unit reads the imaging line-time parameters of the high-resolution camera and the wide-format camera and the image resolution parameters of the two cameras _SH = 10 meters, S _W = 16 meters;

1.2 According to the imaging moment, intercept the area corresponding to I _H in I _W , that is, the reference image I _WR , and the size of I _WR including I _H is (2287 pixels×2500 pixels).

2. Image registration

2.1 Calculate the proportional relationship between the two cameras P _WS = 1.6 from S _W /S _H , and re-sample I _WR by bicubic sampling method to obtain a new reference image I _HR , whose resolution is consistent with I _H ;

2.2 On I _H and I _HR , a total of 9 control points in 3 rows and 3 columns are manually selected evenly, with I _H as a reference, the affine transformation model is adopted

x ₂ =a ₁ +a ₂ ×x ₁ +a ₃ ×y ₁ ,

y ₂ =b ₁ +b ₂ ×x ₁ +b ₃ ×y ₁ ,

Register I _H and I _HR to obtain a reference image I _HR ' that is completely aligned with the spatial position of I _H .

3. Stray light suppression

3.1 Add 0.0001 to the gray value of each pixel in I _HR ', divide each pixel with the same spatial position in I _H and I _HR ', and obtain the proportional coefficient matrix Coff _HW , the elements in the matrix greater than the threshold 5 and less than the threshold 0.0001, make it equal to 1;

3.2 Calculate the image mean values Mean _H = 336.3622 and Mean _HR = 219.0139 of I _H and I _HR ', the ratio reference value Coff _C = Mean _H /Mean _HR = 1.5358, elements greater than 1.5358 in Coff _HW , make its value equal to 1.5358;

3.3 The gray value of each pixel in the reference image I _HR ' is multiplied by the proportional coefficient corresponding to its spatial position in the proportional coefficient matrix Coff _HW , and the difference image Diff=I _H -I _HR ' is calculated, and the gray value of the pixel in Diff is absolutely Elements whose value is greater than the threshold 100 and less than the threshold 10, make it equal to 0;

3.4 Select a square mean value filter template with a side length of 31 to filter the Diff mean value;

3.5 Determine the relationship between the pixel grayscale of the Diff and Sum _Diff = 6.4398×10 ⁷ and the threshold T _destroy = 2×m×n=1.7396×10 ⁷ , if Sum _Diff >T _destroy , go to step 3.6;

3.6 Make a difference between I _H and the gray value of the spatially corresponding pixel in Diff to obtain the stray light suppression result image.

The content that is not described in detail in the description of the present invention belongs to the well-known technology of those skilled in the art.

Claims (3)

1. a satellite image veiling glare eliminates system, it is characterised in that include data acquisition module, image registration module and miscellaneous Astigmatism suppression module；Wherein data acquisition module includes gradation of image acquiring unit and image parameter acquiring unit, image registration Module includes image rough registration unit and image precise registration unit, and veiling glare suppression module includes proportionality coefficient computing unit, difference Value image computing unit and image filtering unit；

Described data acquisition module obtains the high split-phase motor in identical area and wide cut camera image and image from image data base Relevant parameter, and to define wherein piece image be pending image, another piece image is gray reference image；Image registration module With pending image as reference picture, registrate pending image and gray reference image；Veiling glare suppression module is removed pending Veiling glare in image；

Gradation of image acquiring unit obtains pending image and the half-tone information of gray reference image from image data base；Image Parameter acquiring unit obtains pending image and the parameter information of gray reference image, by pending image, gray reference image Half-tone information and parameter information flow to the image rough registration unit in image registration module；

Image rough registration unit obtain data acquisition module provide pending image and the half-tone information of gray reference image and Parameter information, eliminates the differences in resolution between gray reference image and pending image, exports pending image and gray scale ginseng Examine image；Image precise registration unit chooses control point in pending image, chooses aforementioned control point in gray reference image Same place, use affine Transform Model, to gray reference image resampling, obtain the gray scale ginseng with pending image precise registration Examine image, output gray level reference picture and pending image to the proportionality coefficient computing unit of veiling glare suppression module；

Proportionality coefficient computing unit calculates pending image and the ratio of gray reference image correspondence position grey scale pixel value, is compared Example coefficient matrix, and optimized proportion coefficient matrix, the proportionality coefficient Output matrix after optimizing is to error image computing unit；Difference Value image computing unit is by being calculated error image, and error image computing unit obtains the computational methods of error image and is: First it is multiplied by the proportionality coefficient matrix after optimization corresponding with each location of pixels by each grey scale pixel value of gray reference image Proportionality coefficient, obtain the ideal image of pending image rebuild；The ideal image of the pending pending image of figure image subtraction, Obtain error image, error image is flowed to image filtering unit；Error image is done filtering and processes by image filtering unit, To filtered error image；The pending filtered error image of figure image subtraction, obtains veiling glare suppression result images and incites somebody to action This image exports in database.

A kind of satellite image veiling glare the most according to claim 1 eliminates system, it is characterised in that eliminate gray reference figure The method of the differences in resolution between picture and pending image is, according to gray reference figure in the parameter information of gray reference image The resolution ratio of picture obtains the differences in resolution between two sub-pictures divided by the resolution ratio of pending image, with pending image as base Standard, resampling gray reference pattern, make gray reference image consistent with pending image resolution ratio.

3. a satellite image veiling glare removing method, it is characterised in that comprise the following steps:

Step 1, is obtained the high split-phase motor of same area and wide cut camera image and image relevant parameter by data acquisition module, and Defining wherein piece image is pending image, and another width is gray reference image；Data acquisition module includes that gradation of image obtains Take unit and image parameter acquiring unit, gradation of image acquiring unit read high split-phase motor and wide cut camera image I_H、I_W, by Image parameter acquiring unit reads high split-phase motor and parameter during wide cut camera imaging row and two camera image resolution ratio parameters S_H、S_W, Definition high score camera image is pending image, and wide cut camera image is gray reference image；According to imaging moment, intercept gray scale Reference picture I_WIn with pending image I_HCorresponding region, is defined as I_WR, I_WRI need to be comprised_H；

Step 2, by image registration module according to image parameter information, uses the mode of manual reconnaissance with pending image as reference Image, registrates pending image and gray reference image；Image registration module includes image rough registration unit and image precise registration Unit, at image rough registration unit, by S_W/S_HCalculate two image scaled relations P_WS, use the bicubic method of sampling to I_WRHeavily adopt Sample, obtains new gray reference image I_HR, resolution ratio and I_HUnanimously；By image precise registration unit, at I_HAnd I_HRUpper craft is uniform Choose r row c row r × c control point altogether, with I_HFor reference, use affine transformation registration I_HAnd I_HR, obtain and I_HLocus is complete The gray reference image I of full alignment_HR'；

Step 3, according to the spectrum corresponding relation between gray reference image and pending image, the suppression of veiling glare suppression module is treated Process the veiling glare in image, obtain veiling glare suppression result images；Veiling glare suppression module include proportionality coefficient computing unit, Error image computing unit and image filtering unit, in proportionality coefficient computing unit, I_HRIn ', each grey scale pixel value adds 0.0001, I_HAnd I_HR' each pixel that spatial location is identical is divided by, and obtains proportionality coefficient Matrix C off_HW, more than threshold in matrix Value T_CoffHWith less than threshold value T_CoffLElement, make it be equal to 1；I is calculated in proportionality coefficient computing unit_HAnd I_HR' image equal Value Mean_HAnd Mean_HR, ratio reference value Coff_C=Mean_H/Mean_HR, Coff_HWIn more than Coff_CElement, make its value be equal to Coff_C；At error image computing unit, gray reference image I_HRIn ', the gray value of each pixel is multiplied by proportionality coefficient matrix Coff_HWIn the proportionality coefficient corresponding with its locus, calculate error image Diff=I_H-I_HR', in Diff, grey scale pixel value is exhausted To value more than threshold value T_DiffHWith less than threshold value T_DiffLElement, make it be equal to 0；At image filtering unit, selecting the length of side is L's Square mean filter template, to Diff mean filter；By I_HPixel gray value corresponding with space in Diff subtract each other obtain spuious Xanthophyll cycle result images.