Video Object Segmentation Algorithm Based on Background Reconstruction and Its Application
范小聪 FAN Xiao-cong, 李大海 Li Dahai, 王琼华 Wang Qionghua, 刘曦 Liu Xi, 张映权 ZHANG Ying-quan, 刘效勇 LIU Xiao-yong · Chinese Journal of Liquid Crystals and Displays · 2011
A fast video segmentation algorithm based on background reconstruction is proposed for stereo video with static backgrounds.Firstly,the moving foreground areas are found out by the frame difference respectively.And then,the backgrounds shaded by foreground areas are reconstructed.At last,the moving foregrounds could be separated by comparing the backgrounds information with the current frame images and it could achieve the segmentation accurately.After that,two background images reconstructed from the left and right video sequences are matched in the initialization and the parallax image is saved,so that only the foreground parallax images are necessary to be updated.Therefore,the proposed method can improve image compression ratio and reduce storage.Meanwhile,the time for stereo matching is also decreased.