Morphological spatio-temporal simplification for video image segmentation

摘要

In the segmentation-based video compression approach, the number of regions is a basic constraint. This paper presents a spatio-temporal simplification algorithm and a segmentation algorithm in order to segment video images to a small number of regions. The simplification algorithm is based on mathematical morphology. It removes the image details which are perceptually less sensitive for the human visual system and simplifies still objects (null motion) while preserving sensitive moving objects. The segmentation algorithm is based on region growing and region merging, which does not produce artificial boundaries. Experimental results show that, due to the simplification, several still objects may be segmented into one region so that the number of regions is greatly reduced. The segmentation results are well adapted to segmentation-based motion compensation for video compression.