Algorithm for Stereo Disparity after selecting corresponding points in the two images

Hi guys, I'm studying Computer Engineering in Italy and I'm now working for my graduate thesis.

I'm new to Computer Vision, as I've only started reading about it since I began working for my thesis. For now, I've only built a simple application on MATLAB using OpenCV functions (with MEX files built from C++) which computes and displays the Disparity Map between two images using StereoBM, with some sliders to change parameters values.

Now, I've been asked to try to build an application using MATLAB (and, if needed, MEX files from C or C++) which lets the user manually select corresponding points between a pair of stereo images and then computes the disparity map using a Stereo Dense method which takes into account the selected points. It doesn't have to be fast, but I'd like the disparity map to be as good as possible.

Do you guys have any ideas or links that you think could help me in this work?

Answer 1

[Ning Zhang]

There are enough building blocks offered by the Computer Vision System Toolbox, that I believe you don't need to mex the openCV, unless a specific feature detection/descriptor that only offers in OpenCV.

Some examples of the building blocks are:

disparity: https://www.mathworks.com/help/vision/ref/disparity.html

showMatchedFeatures: https://www.mathworks.com/help/vision/ref/showmatchedfeatures.html

with an example: https://www.mathworks.com/help/vision/examples/structure-from-motion-from-two-views.html

You can utilize the APP Designer to build the GUI part. https://www.mathworks.com/products/matlab/app-designer.html