Motion segmentation using region growing and an attention based algorithm

A novel two-stage framework for motion segmentation under stationary background conditions is proposed in this paper. The first stage uses an attention based method to extract motion information. The second stage then applies a region growing technique to motion vectors to obtain motion segmentation and completes motion segmentation with region matching. The algorithm is tested on various video data and experimental results show that the proposed approach can extract detailed motion information from non-rigid objects such as moving people.