Learning to refine depth for robust stereo estimation

作者：

Highlights：

• A multi-branch neural network for disparity estimation is proposed.

• Initial disparity map, confidence measures and disparity gradients can be predicted simultaneously.

• A continuous MRF is proposed to employ the three channel information to refine initial disparity maps.

摘要

•A multi-branch neural network for disparity estimation is proposed.•Initial disparity map, confidence measures and disparity gradients can be predicted simultaneously.•A continuous MRF is proposed to employ the three channel information to refine initial disparity maps.

论文关键词：Stereo matching,Confidence measure,Convolutional neural network

论文评审过程：Received 11 March 2017, Revised 18 July 2017, Accepted 26 July 2017, Available online 31 August 2017, Version of Record 23 October 2017.

论文官网地址：https://doi.org/10.1016/j.patcog.2017.07.027