rectifyStereoImages output looks twisted.

Hi, I have implemented rectifyStereoImages, but the result looks weird. Please look at the images attached. fig1 and 2 are the original files used. Here is my script.

left = webcam(3);
right = webcam(1);
I1 = snapshot(left);
figure(1)
imshow(I1);
I2 = snapshot(right);
figure(2)
imshow(I2);

[J1,J2]=rectifyStereoImages(I1,I2,stereoParams,'OutputView','full');
figure(3)
imshow(cat(3,J1(:,:,1),J2(:,:,2:3)),'InitialMagnification',50);

Answer 1

[Astha Singh]

Hi Sang,

You probably need to fine-tune the 'stereoParams' that you are using.

Most likely that is what is causing the anomaly.

I tried uncalibrated rectification using the two images that you have provided and was able to produce the following rectified image -

You can try this method using the MATLAB Documentation link - Uncalibrated Stereo Image Rectification.

Hope this helps!

[SANG GYU LEE]

Astha! Thank you very much for your comment! My goal is to get the disparity map and the depth (distance) from a camera to an object. I saw the instruction about uncalibrated rectification for stereo vision, but I am not quite sure how https://www.mathworks.com/help/vision/examples/uncalibrated-stereo-image-rectification.html and https://www.mathworks.com/help/vision/examples/depth-estimation-from-stereo-video.html are used together.

So assuming that I have the raw images taken from two cameras, should I do the uncalibrated rectification first -> crop the overlapping parts from the two images -> run stereoCameraCalibrator? If so, doesn't uncalibrated rectification affect the depth value?

Answer 2

[Subhash Variya]

Hello SANG, i have the same problem. do you have the solution now?