Efficient Block-based Motion Segmentation Method using Motion Vector Consistency

This paper presents a new and efficient block-based motion segmentation method based on a novel motion vectors consistency model. In essence, the method utilizes the motion vectors extracted from the video encoding process and groups the associated blocks by region growing technique to achieve motion segmentation. A new motion vector consistency criterion is also introduced to support this block-based region growing mechanism. The proposed method was evaluated by the “Hall Monitor” and “Table Tennis” sequences. Experimental results show that the proposed method can successfully segment the moving objects in the two video sequences, with each object being represented by reasonably good block-based boundary. In addition to that, it is also found that the motion segmentation method can operate together with a video codec to achieve simultaneous real-time video encoding and segmentation, demonstrating its potential in video surveillance applications and region-of-interest video encoding.