1

I am working on finding stereo correspondences using epipolar geometry. I am using p^T * F p'= 0* equality and I know Fundamental Mat's values. Basically , I have 2 rectified images and I am trying to calculate the matches of a point I took from the left image, using the points on the same line in the right image, over this equality. But I did not reach "0" result from this equation. left

right

Input images are rectified ,there is just translation between them and I know Fundamental Matrix

I run ORB detector and flann matcher and I get correspondence with different method. But when I was looking matched points they are not at same line and they do not give 0 result it gave grater than "0" from **p^t F p' equality.What is the reason of this?

What should I do when finding correspondence using epipolar geometry?

If the image is rectified, shouldn't we use the same line to search for correspondence? Do you think it is correct to use this equation?

Burak AKBAŞ
  • 11
  • 1
  • Why not use the existing opencv stereo block matching algorithm ? `cv2.StereoBM_create` or `cv2.StereoSGBM_create`. This does the match along the epipolar line, then you can use the result to output a depth map for example. – Tiphel Jun 28 '21 at 15:49
  • Because I work with epipolar geometry. I want to compare both algorithms – Burak AKBAŞ Jun 28 '21 at 16:18
  • I just gave an answer on a different [question](https://stackoverflow.com/questions/68177733/fundamental-matrix-for-point-correspondence). Looks like a similar problem. Start with normalizing the coordinates and check the method you are using for calculation of F. You will never reach exactly 0. – Grillteller Jun 30 '21 at 09:22

0 Answers0