Gaussian filters are frequently used for scale space based corner detection to remove noise and local variation as many false corners are detected in presence of them. However, an appropriate smoothing scale should be selected for Gaussian filtering m...
Gaussian filters are frequently used for scale space based corner detection to remove noise and local variation as many false corners are detected in presence of them. However, an appropriate smoothing scale should be selected for Gaussian filtering method, which is a difficult task. Moreover, edges are smoothed out in this method, which creates difficulty in corner detection. In this paper, we propose an adaptive filtering method based on the anisotropic diffusion for scale space based corner detectors. A new filtering coefficient was developed. Edges and interior regions were filtered separately by selecting appropriate thresholds. Edges were detected by the Canny edge detector and corners were detected by the Affine Resilient Curvature Scale Space (ARCSS) corner detector. Experimental results demonstrated that the proposed adaptive method can detect more corners in less computational time than that of original ARCSS.