Digital Video Processing, Second Edition

Chapter 5. Video Segmentation and Tracking

Video segmentation refers to partitioning images (frames) or video into spatial, temporal, or spatio-temporal regions that are homogeneous in some feature space. As with any segmentation problem, effective video segmentation requires proper feature selection and an appropriate distance measure. Temporal-segmentation partitions video into shots based on similarity of frames. Spatial-segmentation partitions each video frame into homogeneous regions, which can be achieved by segmenting each frame individually based on color similarity (intra-frame image segmentation). Spatio-temporal segmentation yields temporally connected spatial segments or object trajectories, for example, by inter-frame segmentation ...