Scene identification in image sequence
Yi Tian · Jounal of Xidian University · 2002
Firstly, we detect the points of interest in a pair of images in image sequence, and then match the point set using the gray value differential invariants. Secondly, the fundamental matrix is estimated using an improved Hartleys eight point algorithm. Thirdly, we judge whether the secene between the two images is the same or not by the epipolar geometry constraint. The effectiveness of the proposed algorithm is verified by extensive experiments on real images.