Image lengthening of 'rectifyStereoImages' function application results

Hi. I apply 'rectifyStereoImages' function for rectify stereo image. But results are littlebit weird.

[J1, J2] = rectifyStereoImages(I1, I2, stereoParams, 'OutputView', 'full');

I ran just this code.

I1, I2`s size : 2592x1944, but J1, J2`s size 8704x2378. It`s too long and there is no effect changing the option 'OutputView', 'full' → 'vaild' or reverse.

My stereo camera system is each camera tilts 20 degrees, -20 degrees towards the center (the optical axis draws a V shape), is it common for the output image to be longer when this function is applied?

(first row : input image, second row : output image)

Answer 1

[Song-Hyun Ji]

I am not sure what is your 'steroParams' but they would affect the result. Here is my test code.

ld = load('wideBaselineStereo');

leftImages = imageDatastore(fullfile(toolboxdir('vision'), 'visiondata', ...
    'calibration', 'wideBaseline', 'left'));
rightImages = imageDatastore(fullfile(toolboxdir('vision'), 'visiondata', ...
    'calibration', 'wideBaseline', 'right'));

[imagePoints, boardSize] = detectCheckerboardPoints(leftImages.Files, rightImages.Files);

squareSizeInMillimeters = 29; 
worldPoints = generateCheckerboardPoints(boardSize, squareSizeInMillimeters);

stereoParams = estimateStereoBaseline(imagePoints, worldPoints, ...
    ld.intrinsics1, ld.intrinsics2);
figure; showReprojectionErrors(stereoParams)

I1 = readimage(leftImages, 1);
I2 = readimage(rightImages, 1);
[J1,J2] = rectifyStereoImages(I1,I2,stereoParams,'OutputView','full');
imshow(J1); 

[JONGHAK LEE]

Thank you for your reply. When I ran the your code, I found the same phenomenon. This might be a common phenomenon of the 'rectifyStereoImages()' function.

[Song-Hyun Ji]

In my example, it can create a red-cyan anaglyph to see the stereo effect.

figure; imshow(stereoAnaglyph(J1, J2));

[JONGHAK LEE]

yes, I got same result about my target images. It seems to be due to including black area.

by the way, Are you interested in stereo matching? If so, the code below would be helpful. I have to use SGM or BM, but in my case SGM was not satisfactory, and BM failed completely in any block size.

%% SGM

disparityRange = [0 48];

disparityMap = disparitySGM(J1,J2,"DisparityRange",disparityRange,"UniquenessThreshold",20);

figure

imshow(disparityMap,disparityRange)

title("Disparity Map")

colormap jet

colorbar

%% BM

disparityRange = [0 64];

disparityMap = disparityBM(J1, J2, 'BlockSize', 15, 'DisparityRange', disparityRange);

figure

imshow(disparityMap,disparityRange)

title('Disparity Map')

colormap jet

colorbar

-SGM

-BM

[Song-Hyun Ji]

I think there might be some factors to get the expected result. I made the following example to get the failed result. If you think there is a product problem, please contact the Technical Support team with the reproducible sample files.

close all;
I1 = imread('rectified_left.png');
I2 = imread('rectified_right.png');

A = stereoAnaglyph(I1,I2);
figure;
tiledlayout(3,2)
nexttile([1 2]);
imshow(A)
title('Red-Cyan composite view of the rectified stereo pair image')

J1 = im2gray(I1);
J2 = im2gray(I2);

disparityRange = [0 48];
disparityMap = disparityBM(J1,J2,'DisparityRange',disparityRange,'UniquenessThreshold',20);


nexttile
imshow(disparityMap,disparityRange)
title('Disparity Map BM')
colormap jet
colorbar

disparityRange = [0 48];
disparityMap = disparitySGM(J1,J2,"DisparityRange",disparityRange,"UniquenessThreshold",20);

nexttile
imshow(disparityMap,disparityRange)
title("Disparity Map SGM")
colormap jet
colorbar
%%

I2 = imread('rectified_left.png'); % changed the 2nd image to left camera
I1 = imread('rectified_right.png'); % changed the 1st image to right camera

J1 = im2gray(I1);
J2 = im2gray(I2);

disparityRange = [0 48];
disparityMap = disparityBM(J1,J2,'DisparityRange',disparityRange,'UniquenessThreshold',20);

nexttile
imshow(disparityMap,disparityRange)
title('Disparity Map BM Failed')
colormap jet
colorbar

disparityRange = [0 48];
disparityMap = disparitySGM(J1,J2,"DisparityRange",disparityRange,"UniquenessThreshold",20);

nexttile
imshow(disparityMap,disparityRange)
title("Disparity Map SGM Failed")
colormap jet
colorbar

[JONGHAK LEE]

Hi, Ji. Thanks for your code and I`m pleased that it is helpful for research.

It works in my environment and for the test, I tried to proceed by replacing only input I1, I2 with another images. Even though I just replaced input images, result is similar to the result of I answered earlier, and also It seems similar to the result of 'Failed' in the example you presented.

close all;

I1 = imread('mouse3_left.jpg');

I2 = imread('mouse3_right.jpg');

A = stereoAnaglyph(I1,I2);

figure;

tiledlayout(3,2)

nexttile([1 2]);

imshow(A)

title('Red-Cyan composite view of the rectified stereo pair image')

J1 = im2gray(I1);

J2 = im2gray(I2);

disparityRange = [0 48];

disparityMap = disparityBM(J1,J2,'DisparityRange',disparityRange,'UniquenessThreshold',20);

nexttile

imshow(disparityMap,disparityRange)

title('Disparity Map BM')

colormap jet

colorbar

disparityRange = [0 48];

disparityMap = disparitySGM(J1,J2,"DisparityRange",disparityRange,"UniquenessThreshold",20);

nexttile

imshow(disparityMap,disparityRange)

title("Disparity Map SGM")

colormap jet

colorbar

%%

I2 = imread('mouse2_left.jpg'); % changed the 2nd image to left camera

I1 = imread('mouse2_right.jpg'); % changed the 1st image to right camera

J1 = im2gray(I1);

J2 = im2gray(I2);

disparityRange = [0 48];

disparityMap = disparityBM(J1,J2,'DisparityRange',disparityRange,'UniquenessThreshold',20);

nexttile

imshow(disparityMap,disparityRange)

title('Disparity Map BM Failed')

colormap jet

colorbar

disparityRange = [0 48];

disparityMap = disparitySGM(J1,J2,"DisparityRange",disparityRange,"UniquenessThreshold",20);

nexttile

imshow(disparityMap,disparityRange)

title("Disparity Map SGM Failed")

colormap jet

colorbar

Because there is daily upload limit, I hyperlinked the images what I used 'mouse images'.

Thanks for your help.

[Song-Hyun Ji]

It is necessary to rectify the input images I1 and I2. Rectification ensures that corresponding points in the stereo pair image are aligned along the same rows. To rectify the input stereo pair image, you can utilize the rectifyStereoImages function. It is important to note that the reference image must remain consistent for both rectification and disparity map computation.