Evolving Image Segmentations for the Analysis of Video SequencesAngus A Clark, Barry T Thomas, Evolving Image Segmentations for the Analysis of Video Sequences. Computer Vision and Pattern Recognition 2001, Vol. 2. Anne Jacobs, Thomas Baldwin, (eds.). ISBN 0-7695-1272-0, pp. 290–295. December 2001. PDF, 699 Kbytes.
A methodology for the segmentation of successive frames of a video sequence is presented. Traditional methods, treating each frame in isolation, are computationally expensive, ignore potentially useful information derived from previous frames, and can lead to instabilities in the segmentation over the sequence. The approach developed here, based on the Region Competition algorithm (Zhu and Yuille, IEEE Trans. PAMI, 1996), employs a mesh of active contour primitives, supervised by an MDL energy criterion, to partition the image into homogeneous regions. The inherently dynamic nature of the algorithm allows an initial segmentation to evolve in response to changes observed in the video sequence. Temporal extensions, namely Boundary Momentum, Region Memory, and Optical Boundary Flow, are developed to ease the transition between successive frames. Further enhancements are made by incorporating mechanisms to accommodate the topological discontinuities that can arise during the sequence (e.g. objects entering or leaving the scene). The algorithm is demonstrated using a number of synthetic and real video sequences and is shown to provide an efficient method of segmentation which encourages stability across frames and preserves the quality of the original segmentation over the sequence.