I have a video, and I have applied SIFT and dense optical flow (Farneback) to it (Using python, and open CV). Is there a way to filter the key points given by sift using the dense optical flow, so that I only get the key points of the MOVING objects and it ignores those in the background?
Any help is much appreciated :)
I tried to filter it based on the "projected point" in the next frame, but I was unable to get it to work. At one point I was getting 2 points in the WHOLE video in just random places. I am not sure what I am doing wrong
EDIT --------------
I figured out that the output of the dense optical flow return the "shift" in x and y of every single pixel co-ordinate. I also figured out how to separate the keypoints returned by SIFT by x,y pixel co-ordinate, though I did need to round down or up to turn it into an Int value.
I now need to figure out how the optical flow algorithm decides which points to "color", and then use those co-ordinates to filter the key points. Any ideas?