Abstract:

In order to improve the invariance of the scale invariant feature transform (SIFT) algorithm and reduce mismatches when multiple regions are similar, a novel algorithm for image local invariant feature description is proposed. An improved texture feature descriptor called centersymmetric improved local ternary patterns (CS-ILTP) which is a modified version of the local binary patterns is introduced, and then it is fused with the global distribution of intensity (GDI) as the local feature descriptor in the SIFT. Through iterative transformation, the initial feature point derived from the SIFT converges to affine invariant point and affine invariant region is extracted. Then the image local feature descriptor is obtained by extracting the CS-ILTP and GDI, respectively. The matching experiments show that the proposed algorithm has high robustness and uniqueness. Compared with the SIFT, the number of correct matching feature points in similar regions matching and artificial landmarks matching has increased more than 100% and by 86%, respectively.