Texture Analysis for Shadow Removing in Video-surveillance Systems

The paper presents a new approach for detecting and removing shadows from objects and pedestrians, since shadow removing is a fundamental step in video-surveillance systems for accurate object detection. In order to precisely remove the unwanted shadows, a novel approach is proposed, focused on the problem of representing texture information in terms of redundant systems of functions (frame). The method for discriminating shadows is based on the matching pursuit (MP) algorithm using an over-complete dictionary: the basic idea is to use MP for selecting the best little set of atoms (dictionary functions) of 2D Gabor dictionary and representing texture as linear combination of frame elements. The approach proves how MP is a powerful scheme able to compactly capture detailed textural information of little regions of the image, so MP decomposition coefficients can be used as an exhaustive features in the shadow points detection process. Experimental results validate the algorithm's performance.