Motion segmentation and tracking optimization with edge relaxation in the cellular nonlinear network architecture

We have developed a high-level set of cellular neural net (CNN) functions for finding the segment-borders of moving objects through spatio-temporal relaxation optimization. They are analogic algorithms based on simple CNN instructions considering their implementability in analogic VLSI chips. Motion information extraction from video series is very power consuming. Most computing effort is devoted to motion vector field estimation, object definition and boundary determination. Finding interrelations among small segments, obtained by oversegmentation, involves optimization through merging or separation. In our algorithm the process starts from an oversegmented image, then the segments are merged using information from spatial and temporal auxiliary data: motion fields and motion history. This grouping process is based on neighboring segment similarity in color, speed and time-depth. There is also a feedback to accept or refuse the cancellation of a segment-border. Our parallel approach is independent of the number of segments or objects. We use simple VLSI functions. We develop grouping by stochastic optimization. This relaxation-based motion segmentation can be a basic step of the effective coding of image-series and other automatic motion tracking systems. The proposed system is planned to implement in a cellular nonlinear network chip-set architecture.