Clarification about incorrect image matching points in SIFT
I detected sift key points of two images. (two images have the same scene, but different view angle) after that I did matching of key points. To remove the bad matching I used RANSAC. Most of the matching's are correct. But still I have bad matching. But for research purpose I need to identify why bad matching happens.
By Ransac you mean you estimate the homography transformation with a robust (Ransac) method? Bear in mind that the homography works only with planar scene.
Also, which matching method did you use (nearest neighbor, ...)?
If you want to understand why two keypoints are matched you should display the SIFT histogram and reverse the process like ihog.