omnimvs专题

OmniMVS拜读:End-to-End Learning for Omnidirectional Stereo Matching

图像来源 wide field-of-view (FOV) cameras on an omnidirectional rig 文中具体为4个刚耦合的鱼眼相机 方法 1.将输入的鱼眼图像提取为unary feature maps 通过2D CNN实现 文中使用SegNet+dilated convolution 2.利用feature maps和内外参建立4D feature volum