Unsupervised segmentation of highly dynamic scenes through global optimization of multiscale cues
作者：
Highlights：
• We introduce a global optimization of weighted total variation energy functional.
• The energy combines motion and spectral boundaries with object inside mappings.
• Alternating direction convex optimization provides high-quality salient mapping.
• Integrating mapping with MRF facilitates sequential combination of multiscale cues.
• Feasibility and superiority are demonstrated in segmenting highly dynamic scenes.
摘要
Highlights•We introduce a global optimization of weighted total variation energy functional.•The energy combines motion and spectral boundaries with object inside mappings.•Alternating direction convex optimization provides high-quality salient mapping.•Integrating mapping with MRF facilitates sequential combination of multiscale cues.•Feasibility and superiority are demonstrated in segmenting highly dynamic scenes.
论文关键词：Image sequence segmentation,Dynamic scene,Unsupervised segmentation,Global optimization
论文评审过程：Received 2 July 2014, Revised 25 March 2015, Accepted 20 April 2015, Available online 29 April 2015, Version of Record 16 July 2015.
论文官网地址：https://doi.org/10.1016/j.patcog.2015.04.021
