Online Salient Object Segmentation with Local Weight Assignment
Song Gu, Zheng Ma, Mei Hua Xie · International Journal of Signal Processing Image Processing and Pattern Recognition · 2017
We present a novel on-line algorithm for target segmentation and tracking in video.In our approach, video data is represented by both temporal saliency and spatial one, and segmentation is accomplished by finding the minimum energy label assignment.The pixel-wise weight is assigned for each energy item according to the local information of each feature map.This local weight model enhances the segmentation accuracy.Experiments demonstrates that our approach is effective.