Feature points stereo matching?

edit

Just a comment, you probably have the concept of ROI in stereo matching wrong. It indicates the region around interesting points, which is then pushed through a descriptor for effective representation if I am not mistaken.

Thanks you, Mathieu. But I'd like use stereo feature matching for two reasons. One reason is it will reduce the computation greatly. Another is it will helpful for feature tracking between frames. So if you have any good idea, thanks for sharing.

Dawei there actually are feature matchers in OpenCV. Look at the following link: http://docs.opencv.org/modules/gpu/doc/feature_detection_and_description.html

The bruteforcematching does exactly what you are asking giving two sets of feature points.

Thank you, Steven. I am wondering whether I can run the GPU function on my laptop without GPU.

I guess this wasn't too hard to find out yourself by browsing through the online documentation, but go to this link. http://docs.opencv.org/modules/features2d/doc/features2d.html the link shows you the 2D feature functionality computed on CPU, suggesting different descriptors, feature selectors and possible matchers. Now go try some code ;)

I know opencv has stereo matching method based on region,such as BM,HH,Var ect., dose anybody have any idea on easy way to do the stereo matching based on feature point? Especially, how to apply the epipolar line constraint? Thanks a lots!

@Dawei Jiang !! have you found any solution to your question? i ran into the same problem today