Revisiting graph construction for fast image segmentation
作者：
Highlights：
• A novel way to construct graph for very fast image segmentation with global and local energy functions is proposed.
• Various high-level cues (co-occurrence and saliency) to help build the graphs are developed.
• A fast graph partition optimization objective is proposed.
• A multi-class segmentation method using simple EigenHistograms is pro- posed.
• Extensive experiments on BSDS500, PASCAL VOC, and COCO datasets are conducted to demonstrate the effectiveness of the proposed method.
摘要
•A novel way to construct graph for very fast image segmentation with global and local energy functions is proposed.•Various high-level cues (co-occurrence and saliency) to help build the graphs are developed.•A fast graph partition optimization objective is proposed.•A multi-class segmentation method using simple EigenHistograms is pro- posed.•Extensive experiments on BSDS500, PASCAL VOC, and COCO datasets are conducted to demonstrate the effectiveness of the proposed method.
论文关键词：Image segmentation,Graph partition,Manifold
论文评审过程：Received 21 April 2017, Revised 3 December 2017, Accepted 23 January 2018, Available online 2 February 2018, Version of Record 9 February 2018.
论文官网地址：https://doi.org/10.1016/j.patcog.2018.01.037
