MULTI-SENSOR IMAGE REGISTRATION BASED ON GEOMETRIC AFFINE INVARIANT

A new automatic feature-based registration algorithm of multi-sensor images with affine deformation is presented. The gray scale discrepancy between corresponding pixels and complex geometric deformation are two challenges in the problem of multi-sensor image registration, which make the design of robust feature descriptor very difficult. In the proposed method, the intersections of extracted line segments from both images are used as the feature points, due to their good robustness against gray scale discrepancy. In contrast to most existing feature-based methods that describe feature points directly, we first construct a new feature, named feature segment, based on the feature points, and then describe each feature segment with a descriptor. The additional procedure of feature segment construction enables us to introduce an algebraic affine invariant into the design of feature descriptor. By doing these, the proposed algorithm's robustness to both affine deformation and gray scale discrepancy can be guaranteed. In addition, since the calculation of feature descriptors only involves simple algebraic operations, the proposed method has low computational load. Experimental results using real multi-sensor image pairs are presented to show the merits of the proposed method.