Hi Ahmed, it all depends on what you need. If you have a stereo image pair and the images are undistorted and rectified, you can follow the instruction at the link below

https://it.mathworks.com/matlabcentral/fileexchange/28522-stereo-matching

It is an example of stereo matching and the output is a pixel wise density disparity map.

if the images are not co-registered then you could use the keypoint matching approach. I suggest you to follow the approach at the link below:

https://it.mathworks.com/help/vision/ref/matchfeatures.html

For a more accurate study of the model representing the match between the local keypoint you can apply the ransac method (allowing a fit homography analysis).

Please let me know if any of these suggestions work for your purpose.

Have a good time.

Alessandro

If the images are rectified, the StereoSGBM algorithm in OpenCV is also popular for obtaining stereo disparity maps which can then be turned into 3D points.