Techniques On Interest Points Detection And Image Matching

Techniques on interest points detection based on gray-level image and image matching method using those points are presented in this thesis.Interest points are low-level features where the signal changes two-dimensionally. Firstly, this thesis describes some previous algorithms on interest point detection. Then the performance analysis of those algorithms is given. Finally, the key techniques on interest point detection and factors influencing the performance of the algorithm are detailed from a new point of view.The information available for image matching includes the gray-level of or near interest points and the geometrical relation of interest points. This thesis proposes a new image matching method using gray-level differential invariants and Delaunay Triangulation under the epipolar geometry constraints. Experimental results show good matching in the case of partial visibility and perspective deformation.