Robust Object Detection and Segmentation Based on Radial Reach Correlation

We propose a novel algorithm for robust object detection and segmentation, which is based on a new robust dissimilarity measure called as Radial Reach Correlation (RRC). The capability of detecting new objects from a complex background is one of the most (,I. r ~ t r e n 2 :%qe (D c-~rren: scerle (c) RRC imaqe important technologie for many vision systems. The usual approach for detecting new objects from a background is simple background subtraction. However, it is strongly influenced of brightness changes such as falge positives shadows and gain change of the imaging system. The RRC is a new robust dissimilarity measure and has 'd's?le '' t k round (e) s y l e h i k round ( f ) s le back round sub ract lonqth-8) sub ractlonqth-15) s s r a c t l o n ? t h 3 0 ) a well-formed probabilistic model of binary or normal density. The RRC evaluates the local texture between Figure 1: Overview. a background image and the current scene and realize robust object detection under poor conditions. Experflows [7] have been proposed. However, it is well known imental results with real images show the effectiveness that such techniques have problems in stability and calof the proposed method. culation costs. In this paper, we propose a novel algorithm for ro