LSI: Latent semantic inference for natural image segmentation

作者：

Highlights：

• We do latent semantic inference for natural image segmentation with Cross-R&S model.

• We formulate segmentation problem as label inference problem.

• We design the LSI to capture the context, layout, and scale information of pixels.

• Cross-Region and Cross-Scale potentials are designed in the LSI framework.

摘要

Highlights•We do latent semantic inference for natural image segmentation with Cross-R&S model.•We formulate segmentation problem as label inference problem.•We design the LSI to capture the context, layout, and scale information of pixels.•Cross-Region and Cross-Scale potentials are designed in the LSI framework.

论文关键词：Image Segmentation,MRF Model,Label Inference

论文评审过程：Received 26 August 2015, Revised 14 February 2016, Accepted 3 March 2016, Available online 13 March 2016, Version of Record 23 August 2016.

论文官网地址：https://doi.org/10.1016/j.patcog.2016.03.005