A fast automatic VOP generation using boundary block segmentation

• Venue: Real-Time Imaging
• Year: 2004

With the increase of multimedia applications and content-based functionalities, efficient video coding methods are necessary. The moving picture experts group-4 (MPEG-4) provides content-based functionalities by introducing the concept of the video object plane (VOP). Each frame of the input sequence is segmented into a number of arbitrarily shaped image regions or VOP's so that each VOP describes a semantically meaningful object or video content of interest. For real-time applications of MPEG-4. a fast automatic object segmentation method of video sequences is needed. We propose a fast automatic VOP generation algorithm composed of two parts: object block segmentation and boundary block segmentation. The former defines block-based object regions in a frame and the latter generates the pixel-based object mask. Block-based object region search and restriction of segmentation regions to boundary blocks reduce a computational load. Experimental results with two test sequences show the effectiveness of the proposed algorithm in terms of the visual quality of segmentation results and the computation time.