A three-stage on-track image strip inner step removal method

Abstract

The present invention relates to a three-segment on-track image de-stairing processing method, which belongs to the field of image processing technology and solves the problem of steps appearing after the strip images are spliced ​​and the image block effect caused by different image stretching degrees after the splicing of different strip images. The method first divides the strip image to be corrected into four regions, namely, a retention region, a sin region, a linear region and a vignetting region; according to the position of the inflection point of the image to be corrected, different stretching coefficients are constructed for the four regions, and the corresponding regions are stretched using different stretching coefficients, which can effectively remove the phenomenon of steps in the strip and avoid the block effect. The method of the present invention adaptively solves the stretching coefficient according to the characteristics of the image itself, so that the image transitions evenly and avoids the stair effect.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a three-segment on-track image strip inner step removal method. Background Art

[0002] In actual engineering, when the visible light remote sensing camera is imaging on-orbit, a staircase phenomenon appears in the strips. Conventional fusion methods do not take into account the vignetting area caused by the mechanical stitching size used in the camera design process. After the same scene is stitched, there will be inconsistencies between the strips, affecting the visual effect; and different strips of images will have different stretching degrees after stitching, which will bring about image block effects. Summary of the invention

[0003] In order to solve the above problem, the present invention proposes a three-stage on-track image strip intra-step removal method.

[0004] A three-stage on-track image strip inner step removal method, the method comprising the following steps:

[0005] Step 1: Mark the N strip images to be spliced ​​from left to right, and mark the i-th strip image as Img i , where i=1-N, and the strip image is a CCD or CMOS image after radiation correction;

[0006] Step 2: Obtain the size of the optical vignetting area of ​​N stripe images respectively, wherein the size of the optical vignetting area between the i-th stripe image and the i+1-th stripe image is marked as W i ;

[0007] Step 3: Obtain the i+1th strip image Img by statistically analyzing the image regularity i+1 The left and right turning point columns are denoted as left_point and right_point respectively;

[0008] Step 4: Take the i+1th strip image Img i+1 It is divided into four areas, namely the holding area, sin area, linear area and vignetting area. The column coordinate ranges of each area are [1, left_point-1], [left_point, right_point], [right_piont+1, col i+1 -W i +2], [col i+1 -W i +3,col i+1 ], where col i+1 For the strip image Img i+1 Width;

[0009] Step 5: Take the i-th strip image Img i As the reference standard, correct the i+1th strip image Img i+1 The image correction process is shown in the following steps (1) to (14):

[0010] Step (1): Divide the i-th strip image Img i The reference area is [left_pos-W i / 2,col i ], where col i For the strip image Img i width, let left_pos = col i ;

[0011] Step (2): Calculate the position center_pos = W i / 2;

[0012] Step (3): Calculate the left reference image Img i The grayscale mean of the left_pos-center_pos column in the local area is calculated as follows:

[0013] Left_img_value=mean(Img i (:,left_pos-center_pos))

[0014] Among them, mean is the symbol for finding the mean of the image, Img i (:,left_pos-center_pos) represents the image Img i The pixels of all rows of the left_pos-center_pos column, which is a column vector;

[0015] Step (4): Calculate the right strip image Img i+1 The grayscale mean of the right_pos column of i / 2, the calculation formula is as follows:

[0016] Right_img_value=mean(Img i+1 (:,right_pos))

[0017] Among them, mean is the symbol for finding the mean of the image, Img i+1 (:, right_pos) is the image Img i+1 The pixels of all rows of the right_pos column, which is a column vector;

[0018] Step (5): Two proportional coefficients k are calculated by calculating the grayscale mean ratio of the overlapping vignetting areas of the left and right images. 300 and k1, the calculation formulas are as follows:

[0019] k 300 =Right_img_value*m / (Left_img_value+Right_img_value)

[0020] k1=Right_img_value / Left_img_value

[0021] Where m is the proportionality coefficient;

[0022] Step (6): Construct the proportional coefficients of all points in the linear region: K_1_300 = Area_x1.*Step1+k1, where the column vector Area_x1 = 1:1:(left_point-1), Step1 = (k 300 -k1) / (left_point-1), ".*" represents dot product;

[0023] Step (7): Calculate the inflection point length as Len_x=right_point-left_point+1, and construct the vector x=0:1:Len_x-1;

[0024] Step (8): Calculate the linear region initialization coefficient K0 = [0:1:Len_x-1].*Step2+k1, where Step2 = (1-k 300 ) / Len_x;

[0025] Step (9): Calculate the linear region coefficient K according to the linear region initialization coefficient K0, the calculation method is: K = fliplr (K0), where fliplr is the left-right flip function in the matlab software;

[0026] Step (10): The proportional coefficients at both ends of the linear region are the head end coefficient K head = K(1) and terminal coefficient K tail =K(Len_x);

[0027] Step (11): Construct the stretching function of the sin region:

[0028] K_sin=1-(K tail -K head ) / n*sin(x*2*π / (Len_x-1)), where n is the adjustment coefficient;

[0029] Step (12): Calculate parameter K_sin_sub=K_sin.*K;

[0030] Step (13): Construct the i+1th strip image Img i+1 The different stretching coefficients K_sin_new for the four regions are constructed as follows:

[0031] a. For the retention area, the column coordinate range is 1 to left_point-1, and the stretching coefficient K_sin_new(1,1:left_point-1)=1;

[0032] b. For the sin area, the column coordinate range is from left_point to right_point, and the stretching coefficient

[0033] K_sin_new(1,left_point:right_point)=K_sin_sub;

[0034] c. For the linear region, the column coordinate range is from right_point+1 to col i+1 -W i +2, stretch coefficient K_sin_new(1,right_point+1:col i+1 -W i +2) = k 300 ;

[0035] d. For the vignetting area, the column coordinate range is col i+1 -W i +3 to col i+1 , stretch coefficient K_sin_new(1,col i+1 -W i +3:col i+1 ) = K_1_300;

[0036] Step (14): Expand the stretch coefficient K_sin_new into a matrix K_new, whose size is the same as the image Img i+1 Consistent, each row of the matrix K_new is assigned the value K_sin_new, then the i+1th strip image Img i+1 The image after correction is: JzImg i+1 =Img i+1 .*K_new.

[0037] Beneficial effects of the present invention:

[0038] The present invention relates to a three-stage on-track image de-stairing processing method, which first divides the strip image to be corrected into four regions, namely, a holding region, a sin region, a linear region, and a vignetting region; according to the position of the inflection point of the image to be corrected, different stretching coefficients are constructed for the four regions, and the corresponding regions are stretched using different stretching coefficients, which can effectively remove the stair phenomenon in the strip and avoid the block effect. The method of the present invention adaptively solves the stretching coefficient according to the characteristics of the image itself, so that the image transitions evenly and avoids the stair effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a flow chart of a three-stage on-track image strip intra-step removal method according to one embodiment of the present invention;

[0040] Figure 2 Schematic diagram of image area division;

[0041] Figure 3 It is a CCD step stripe image;

[0042] Figure 4 for Figure 3 The image shown is an image after the steps within the strip are removed by the three-stage on-track image strip step removal method of the present invention. DETAILED DESCRIPTION

[0043] The technical solution of the present invention will be described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0044] In one embodiment, if Figure 1 As shown, the present invention provides a three-stage on-track image strip inner step removal method, which specifically includes the following steps:

[0045] Step 1: Mark the N strip images to be spliced ​​from left to right, and mark the i-th strip image as Img i , where i = 1 to N, and the strip image to be spliced ​​refers to the CCD or CMOS image after radiation correction;

[0046] Step 2: Obtain the size of the optical vignetting area of ​​N stripe images respectively, wherein the size of the optical vignetting area between the i-th stripe image and the i+1-th stripe image is marked as W i ;

[0047] Step 3: Obtain the i+1th strip image Img by statistically analyzing the image regularity i+1 The left and right turning point columns are denoted as left_point and right_point respectively;

[0048] Step 4: Image segmentation

[0049] The i+1th strip image Img i+1 Divided into four areas, such as Figure 2 As shown, the first area is the holding area, the second area is the sin area, the third area is the linear area, and the fourth area is the vignetting area. The column coordinate ranges of the holding area, sin area, linear area, and vignetting area are [1, left_point-1], [left_point, right_point], [right_piont+1, col i+1 -W i +2], [col i+1 -W i +3,col i+1 ], where col i+1 For the strip image Img i+1 The width is the maximum value of the column coordinate;

[0050] Step 5: Construct the stretch factor and correct the image according to the stretch factor

[0051] Take the i-th strip image Img i As the reference standard, correct the i+1th strip image Img i+1 The image correction process is shown in the following steps (1) to (14):

[0052] Step (1): Divide the i-th strip image Img i The reference area is [left_pos-W i / 2,col i ], where col i For the strip image Img i The width is the maximum value of the column coordinate, let left_pos = col i ;

[0053] Step (2): Calculate the position center_pos = W i / 2;

[0054] Step (3): Calculate the left reference image Img i The grayscale mean of the left_pos-center_pos column in the local area is calculated as follows:

[0055] Left_img_value=mean(Img i (:,left_pos-center_pos))

[0056] Among them, mean is the symbol for finding the mean of the image, Img i (:,left_pos-center_pos) represents the image Imgi The pixels of all rows of the left_pos-center_pos column, which is a column vector;

[0057] Step (4): Calculate the right strip image Img i+1 The grayscale mean of the right_pos column of i / 2, the calculation formula is as follows:

[0058] Right_img_value=mean(Img i+1 (:,right_pos))

[0059] Among them, mean is the symbol for finding the mean of the image, Img i+1 (:, right_pos) is the image Img i+1 The pixels of all rows in the right_pos column, which is also a column vector;

[0060] Step (5): Two proportional coefficients k are calculated by calculating the grayscale mean ratio of the overlapping vignetting areas of the left and right images. 300 and k1, the calculation formulas are as follows:

[0061] k 300 =Right_img_value*m / (Left_img_value+Right_img_value)

[0062] k1=Right_img_value / Left_img_value

[0063] Where m is the proportionality coefficient, and “*” represents the product between two values;

[0064] Step (6): Construct the proportional coefficients of all points in the linear region: K_1_300 = Area_x1.*Step1+k1, where Area_x1 is a column vector, Area_x1 = 1:1:x1, which means that the elements in the column vector Area_x1 are all values ​​from 1 to x1 with a step length of 1; Step1 = (k 300 -k1) / x1; where x1=left_point-1, ".*" represents the dot product in MATLAB software. The dot product in MATLAB is used to indicate that each data in a matrix or vector is multiplied by a coefficient;

[0065] Step (7): Calculate the inflection point length as Len_x=right_point-left_point+1, and construct the vector x=0:1:Len_x-1;

[0066] Step (8): Calculate the linear region initialization coefficient K0 = [0:1:Len_x-1].*Step2+k1, where Step2 = (1-k 300 ) / Len_x;

[0067] Step (9): Calculate the linear region coefficient K according to the linear region initialization coefficient K0, and the calculation method is: K = fliplr (K0), wherein fliplr is the left-right flip function in the matlab software, and fliplr (K0) represents the center-symmetric flipping of the linear region initialization coefficient K0;

[0068] Step (10): The proportional coefficients at both ends of the linear region are the head end coefficients K head = K(1) and terminal coefficient K tail =K(Len_x), where K(1) represents the function value of the linear region coefficient K when the column coordinate is 1, and K(Len_x) represents the function value of the linear region coefficient K when the column coordinate is Len_x;

[0069] Step (11): Construct the stretching function of the sin region:

[0070] K_sin=1-(K tail -K head ) / n*sin(x*2*π / (Len_x-1)), where n is the adjustment coefficient, the range of n is 0.5 to 1, "*" represents the product of values, and sin(x*2*π / (Len_x-1)) is a trigonometric function;

[0071] Step (12): Calculate parameter K_sin_sub=K_sin.*K;

[0072] Step (13): Construct the i+1th strip image Img i+1 The different stretching coefficients K_sin_new for the four regions are constructed as follows:

[0073] a. For the hold area, the column coordinate range is 1 to left_point-1, and the stretch coefficient K_sin_new is (1,1:left_point-1)=1;

[0074] b. For the sin area, the column coordinate range is from left_point to right_point, and the stretching coefficient

[0075] K_sin_new(1,left_point:right_point)=K_sin_sub;

[0076] c. For the linear region, the column coordinate range is from right_point+1 to col i+1 -W i +2, stretch coefficient K_sin_new(1,right_point+1:col i+1 -W i +2) = k 300 ;

[0077] d. For the vignetting area, the column coordinate range is col i+1 -W i +3 to col i+1 , stretch coefficient K_sin_new(1,col i+1 -W i +3:col i+1 ) = K_1_300;

[0078] Step (14): The stretching coefficient K_sin_new is a row vector, which is expanded into a matrix K_new. The size of the matrix K_new is the same as that of the image Img i+1 Consistent, each row of the matrix K_new is assigned the value K_sin_new, then the i+1th strip image Img i+1 The image after correction is: JzImg i+1 =Img i+1 .*K_new.

[0079] The three-stage on-track image de-stair processing method proposed in this embodiment first divides the strip image to be corrected into four regions, namely, the holding region, the sin region, the linear region, and the vignetting region; according to the position of the inflection point of the image to be corrected, different stretching coefficients are constructed for the four regions, and then the corresponding regions are stretched using different stretching coefficients, which can effectively remove the stair phenomenon in the strip and avoid the block effect. The method of this embodiment adaptively solves the stretching coefficient according to the characteristics of the image itself, so that the image transitions evenly and avoids the stair effect.

[0080] A specific example is given below to further illustrate the technical solution of the present invention.

[0081] Step 1: Based on the input N radiation-corrected CCD or CMOS strip images to be stitched, the i-th image from left to right is marked as Img i , where i = 1 to 11. In this example, a CCD step stripe image is taken as an example.

[0082] Step 2: Obtain the optical vignetting area sizes of the 11 strip images respectively, wherein the vignetting area size between the first image and the second image is marked as W1=48.

[0083] Step 3: By statistically analyzing the image regularity, the left and right inflection point column positions of the second image Img2 are obtained, which are left_point = 2353 and right_point = 3613. The inflection point refers to the position where the grayscale mutation occurs in the flat area within the CMOS chip.

[0084] Step 4: In this step, the second image Img2 is first divided into four regions, marked as the holding region (column coordinate range is [1, x1]), the sin region (column coordinate range is [x1+1, x2]), the linear region (column coordinate range is [x2+1, col2-W1+2]) and the vignetting region (column coordinate range is [col2-W1+3: col2]), where x1=left_point-1=2352; x2 is the column coordinate of the image Img2, and its value is right_point; col2 is the width of the image img2, which is 4096, that is, the maximum value of the column coordinate.

[0085] Step 5: Here, we take i=1 as an example, that is, we take the first image Img1 as the reference standard and correct the second image Img2. The second image Img2 before correction is as follows: Figure 3 shown.

[0086] Taking the first image Img1 as a reference standard, the correction image processing process of correcting the second image Img2 specifically includes the following steps:

[0087] Step (1): Divide the reference area of ​​the first image W1 into [left_pos-W1 / 2,col1], where col1 is the width of image W1, i.e. the maximum value of the column coordinate. Let left_pos = 4096;

[0088] Step (2): calculate the position center_pos=48 / 2=24;

[0089] Step (3): Calculate the grayscale mean of the left_pos-center_pos=4072th column of the local area of ​​the left reference image Img1. The calculation formula is as follows: Left_img_value=mean(Img1(:,4072)); where mean represents the mean of the position, and Img1(:,4072) represents the pixels of all rows of the 4072th column of the image Img1, which is a column vector.

[0090] Step (4): Calculate the mean of the right_pos of the right image Img2, right_pos=24, and the calculation formula is as follows:

[0091] Right_img_value=mean(Img2(:,24))

[0092] Step (5): Calculate two proportionality coefficients k 300 and k1:

[0093] k 300 =Right_img_value*m / (Left_img_value+Right_img_value)

[0094] k1=Right_img_value / Left_img_value

[0095] Step (6): Construct the proportional coefficients of all points in the linear region:

[0096] K_1_300=Area_x1.*Step1+k1

[0097] Among them, Area_x1=1:1:x1, Step1=(k 300 -k1) / x1, x1=left_point-1;

[0098] Step (7): Calculate the inflection point length according to the left and right inflection point column coordinates of the second image Img2:

[0099] Len_x=right_point-left_point+1=1261

[0100] The construction vector is x=0:1:1260.

[0101] Step (8): Calculate the linear region initialization coefficient K0 = [0:1:Len_x-1].*Step2+k1, where Step2 = (1-k 300 ) / Len_x;

[0102] Step (9): Calculate the linear region coefficient K = fliplr (K0), where fliplr is the left-right flip function in the Matlab software.

[0103] Step (10): The proportional coefficients at both ends of the linear region are the head end coefficients K head = K(1) and terminal coefficient K tail =K(Len_x);

[0104] Step (11): Construct the stretching function of the sin region:

[0105] K_sin=1-(K tail -K head ) / 0.6*sin(x*2*π / (Len_x-1));

[0106] Step (12): Calculate parameter K_sin_sub=K_sin.*K;

[0107] Step (13): Construct different stretching coefficients K_sin_new for the four regions of the second image Img2:

[0108] Keep area: K_sin_new(1,1:2352)=1;

[0109] sin area: K_sin_new(1,2353:3613)=K_sin_sub;

[0110] Linear region: K_sin_new(1,3614:4050)=k 300 ;

[0111] Vignetting area: K_sin_new(1,4051:4096)=K_1_300;

[0112] Step (14): The image of the second image Img2 after correction is: JzImg2=Img2.*K_sin_new.

[0113] The above embodiment is a process of correcting the second image with the first image as a reference. The correction principles of other images are the same and will not be described in detail here. Figure 4 Shown Figure 3 The image after removing the steps in the strip by the three-stage on-track image strip inner step removal method of the present invention is obtained by Figure 3 and Figure 4 It can be seen from the comparison that the present invention can effectively remove the stair-step phenomenon within the strip, avoid the image block effect, and improve the image processing quality.

[0114] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0115] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A three-stage on-track image strip inner step removal method, characterized in that: The method comprises the following steps: Step 1: Mark the N strip images to be spliced ​​from left to right, and mark the i-th strip image as Img i , where i=1-N, and the strip image is a CCD or CMOS image after radiation correction; Step 2: Obtain the size of the optical vignetting area of ​​N stripe images respectively, wherein the size of the optical vignetting area between the i-th stripe image and the i+1-th stripe image is marked as W i ; Step 3: Obtain the i+1th strip image Img by statistically analyzing the image regularity i+1 The left and right inflection points are marked as left_point and right_point respectively. The inflection point refers to the position where a grayscale mutation occurs in the flat area of ​​the CCD or CMOS chip. Step 4: Take the i+1th strip image Img i+1 It is divided into four areas, namely the holding area, sin area, linear area and vignetting area. The column coordinate ranges of each area are [1, left_point-1], [left_point, right_point], [right_piont+1, col i+1 -W i +2], [col i+1 -W i +3,col i+1 ], where col i+1 For the strip image Img i+1 Width; Step 5: Take the i-th strip image Img i As the reference standard, correct the i+1th strip image Img i+1 The image correction process is shown in the following steps (1) to (14): Step (1): Divide the i-th strip image Img i The reference area is [left_pos-W i / 2,col i ], where col i For the strip image Img i width, let left_pos = col i ; Step (2): Calculate the position center_pos = W i / 2; Step (3): Calculate the left reference image Img i The grayscale mean of the left_pos-center_pos column in the local area is calculated as follows: Left_img_value=mean(Img i (:,left_pos-center_pos)) Among them, mean is the symbol for finding the mean of the image, Img i (:,left_pos-center_pos) represents the image Img i The pixels of all rows of the left_pos-center_pos column, which is a column vector; Step (4): Calculate the right strip image Img i+1 The grayscale mean of the right_pos column of i / 2, the calculation formula is as follows: Right_img_value=mean(Img i+1 (:,right_pos)) Among them, mean is the symbol for finding the mean of the image, Img i+1 (:, right_pos) is the image Img i+1 The pixels of all rows of the right_pos column, which is a column vector; Step (5): Two proportional coefficients k are calculated by calculating the grayscale mean ratio of the overlapping vignetting areas of the left and right images. 300 and k1, the calculation formulas are as follows: k 300 =Right_img_value*m / (Left_img_value+Right_img_value) k1=Right_img_value / Left_img_value Where m is the proportionality coefficient; Step (6): Construct the proportional coefficients of all points in the linear region: K_1_300 = Area_x1.*Step1+k1, where the column vector Area_x1 = 1:1:(left_point-1), Step1 = (k 300 -k1) / (left_point-1), ".*" represents dot product; Step (7): Calculate the inflection point length as Len_x=right_point-left_point+1, and construct the vector x=0:1:Len_x-1; Step (8): Calculate the linear region initialization coefficient K0 = [0:1:Len_x-1].*Step2+k1, where Step2 = (1-k 300 ) / Len_x; Step (9): Calculate the linear region coefficient K according to the linear region initialization coefficient K0, the calculation method is: K = fliplr (K0), where fliplr is the left-right flip function in the matlab software; Step (10): The proportional coefficients at both ends of the linear region are the head end coefficient K head = K(1) and terminal coefficient K tail = K(Len_x); Step (11): Construct the stretching function of the sin region: K_sin=1-(K tail -K head ) / n*sin(x*2*π / (Len_x-1)), where n is the adjustment coefficient; Step (12): Calculate parameter K_sin_sub=K_sin.*K; Step (13): Construct the i+1th strip image Img i+1 The different stretching coefficients K_sin_new for the four regions are constructed as follows: a. For the retention area, the column coordinate range is 1 to left_point-1, and the stretching coefficient K_sin_new(1,1:left_point-1)=1; b. For the sin area, the column coordinate range is from left_point to right_point, and the stretching coefficient K_sin_new(1,left_point:right_point)=K_sin_sub; c. For the linear region, the column coordinate range is from right_point+1 to col i+1 -W i +2, stretch coefficient K_sin_new(1,right_point+1:col i+1 -W i +2) = k 300 ; d. For the vignetting area, the column coordinate range is col i+1 -W i +3 to col i+1 , stretch coefficient K_sin_new(1,col i+1 -W i +3:col i+1 ) = K_1_300; Step (14): Expand the stretch coefficient K_sin_new into a matrix K_new, whose size is the same as the image Img i+1 Consistent, each row of the matrix K_new is assigned the value K_sin_new, then the i+1th strip image Img i+1 The image after correction is: JzImg i+1 =Img i+1 .*K_new.

2. A three-stage on-track image strip inner step removal method according to claim 1, characterized in that: The adjustment coefficient n ranges from 0.5 to 1.

Images