Generic Feature Extraction for Image/Video Analysis

This paper presents a generic algorithm for extracting invariant salient features from images and video sequences. The detected features/regions are invariant under viewpoint variations and illumination changes. Repeatability rate is used for evaluating the invariance of the feature detector. We also apply our feature detector on object recognition experiments. Experimental results show our proposed algorithm is very effective for recognizing objects in images/video with viewpoint changes and background variations.