A video segmentation algorithm based on spatial-temporal information

In the MPEG-4 coding standard, each frame of a video sequence is coded based on the video object planes (VOPs). Video segmentation to yield VOPs plays an important role before coding and is the basis of object-based coding. The paper presents a video segmentation algorithm based on spatial-temporal information. In temporal segmentation, a fast motion estimation algorithm based on the prediction of a neighbouring block's motion vector is proposed, and in spatial segmentation, an algorithm based on the re-quantization of the intensity of an image is proposed. Then, the combination of the temporal segmentation and the spatial segmentation produces the VOPs. Experimental results show that it is a robust and fast algorithm with better results.