Omnidirectional stereo depth estimation based on spherical deep network

作者：

Highlights：

• The epipolar lines of omnidirectional stereo are the great circles on sphere.

• The Mesh CNNs can extract features on sphere efficiently.

• The spherical epipolar constraint can improve the omnidirectional depth estimation.

• Multi-stage cascade learning is helpful for depth estimation.

摘要

•The epipolar lines of omnidirectional stereo are the great circles on sphere.•The Mesh CNNs can extract features on sphere efficiently.•The spherical epipolar constraint can improve the omnidirectional depth estimation.•Multi-stage cascade learning is helpful for depth estimation.

论文关键词：Omnidirectional depth estimation,Spherical convolutional network,Stereo matching,Cascade learning

论文评审过程：Received 6 June 2021, Revised 26 July 2021, Accepted 30 July 2021, Available online 2 August 2021, Version of Record 23 August 2021.

论文官网地址：https://doi.org/10.1016/j.imavis.2021.104264