Contourlet Based Interest Points Detector

This paper proposes a robust algorithm for detecting interest points based on the nonsubsampled contourlet transform (NSCT). The NSCT provides multiscale decomposition with directional filters at each scale. Furthermore, NSCT is very efficient in extracting the geometric information of images and therefore it has very good feature localization. The NSCT-based point detector is compared to the widely used Harris and difference of Gaussian (DoG) interest point detectors. The experimental results reveal the robustness of the proposed algorithm to rotation, scale and viewpoint changes.