การแยกส่วนภาพวัตถุสำหรับการเข้ารหัสสัญญาณวิดีโอ บนพื้นฐานของคอนเทนต์ตามข้อกำหนดของ MPEG-4 / ดัชกรณ์ ตันเจริญ = Object segmentation for content based video coding according to MPEG-4 / Datchakorn Tancharoen
To propose a video object segmentation algorithm according to the MPEG-4 which provides the object based video coding standard. This algorithm is based on the combination of multiple features including spatial and temporal informations. The spatial segmentation includes image simplification, boundary detection and region assignment by the watershed algorithm. The morphological opening and closing are used in image simplification process, subsequently, the multiscale gradient approximation is used in boundary detection. The proposed algorithm can reduce the number of regions and uncertain pixels in order to reduce the computational load in the watershed process. By the way, the temporal segmentation using change detection and morphological operator can create the change mask. In addition, the decision process and the memory of the previous segmentation result are used to determine object boundary and detect object tracking. Exploiting the proposed algorithm to segment the moving object in the video conferencing sequence, the subjective test showed meaningful and good quality video objects compared with those of COST211 results. Moreover, the objective test showed an error of segmentation result when compared with reference mask, which is less than 15% of the object size, while the miss of moving object is less than 5% of the object size. Therefore, this algorithm can provide the semantic meaningful video objects and preserve the content based functionalities according to MPEG-4.
