Image matching by topological relations and geometric features

The aim of the present work was to solve the problem that global and local features cannot be considered comprehensively in image matching, with initial feature points in the images extracted by SURF algorithm and triangulated later. Considering topological relations of feature points in the mesh, we construct
feature vector of each point using cross rations of each edge, i.e. geometric invariance. Image correspondences are acquired by these vectors. Combination of topological relations and geometric features can overcome defects of traditional methods effectively. The method is effective for images with blur, rotation and noise.