I have a pair of close to rectified images that I want to produce a disparity map out of. However, I am not getting good results (cvFindStereoCorrespondence matching very few points, black disparity map, etc.) Anyone know why this might be?
Does the cvFindStereoCorrespondence function strictly work only for 100% completely rectified images? (Because of noise in the images and the calibration process I have done, I am not able to get completely rectified images.)
What are the optimal parameters for CvStereoBMState that I can use to get the best results? (Time isn't an issue.)