Diffusive likelihood for interactive image segmentation
作者：
Highlights：
• Diffusive likelihood strategy is proposed to obtain accurate estimation of prior probability from limited seeds.
• Superpixel-based grouping cues are introduced to enforce continuity for the object extraction.
• We construct the segmentation model by combining the geometrical adjacency and long range grouping cues.
• A joint optimization technique is utilized to solve a pair of sub-module functions
摘要
•Diffusive likelihood strategy is proposed to obtain accurate estimation of prior probability from limited seeds.•Superpixel-based grouping cues are introduced to enforce continuity for the object extraction.•We construct the segmentation model by combining the geometrical adjacency and long range grouping cues.•A joint optimization technique is utilized to solve a pair of sub-module functions
论文关键词：Interactive image segmentation,Likelihood diffusion,Perceptual learning,Graph cuts
论文评审过程：Received 4 September 2017, Revised 22 January 2018, Accepted 18 February 2018, Available online 19 February 2018, Version of Record 2 March 2018.
论文官网地址：https://doi.org/10.1016/j.patcog.2018.02.023
