Fast tracking of semantic video object based on motion prediction and subregion extraction

Extraction quality and speed are two fundamental problems for semantic video object extraction. This paper introduces a novel fast-speed tracking algorithm for extracting semantic video object in image sequences. This approach first extracts the subregion that covers the contour of a semantic video object, by using motion prediction and mathematical morphology operators to generate the inner and outer contour of the subregion, and then an optimized tracking scheme is employed to track this particular subregion instead of the whole image. Results on real sequences show this method greatly improves the processing speed compared to the earlier tracking algorithms and that, keeps the quality of the extracted object. Therefore, it is a promising approach in real time processing system based on video object.