A Multi-scale Visual Salient Feature Points Extraction Method Based on Gabor Wavelets

Keypoints are important features in the perceptual system of humans. They provide important information for focus-of-attention (FoA) and object categorization/recognition. Different types of keypoints have been used in computer vision applications. In this paper, we propose a method which extracts "salient" points in the meaning of biological vision by utilizing the multi-scale Gabor energy operator. The resulting operator responds strongly to corners, isolated lines, edges, and contours. And with the increase of scale, the extracted keypoints tend to illustrate important structures of the image. We show that the Gabor energy map provides very useful information of a saliency map for FoA and object recognition.