A pixel-parallel moving object segmentation and tracking algorithm for video surveillance applications

In this paper, the design and on-chip implementation of an active contour technique for segmenting and tracking moving objects is proposed. The module is intended to be part of a full-automatic real time video surveillance system. All the processing steps, including the active contours are fit to be performed on a cellular vision system with focal plane computing capabilities. Therefore, a fast time of response is obtained, leaving room for the integration of the subsequent high-level processing steps. The algorithm has been fully implemented on an EyeRIS Vision System. Some experimental results and data from the on-chip implementation are included.