Comparison of the feature point tracking method in image sequences

In order to reconstruct 3D structure from image sequences we have to do tracking feature points on the sequence. There are some methods to track feature points on image sequence. One is to track the feature points and remove outliers by the geometric constraints between successive images. Another method is a camera optical system approximated by the affine camera model and tracking feature points. In this paper we took some image sequences and compared the results of experiments of the feature point tracking. Consequently, we obtained correct correspondences with high probability in methods by using geometric constraint. But the number of correspondences was small. On the other hand, we obtained many correct correspondences and some incorrect correspondences in method by using the affine space constraint.