[Stereo_unsupervised][cvpr16]Unsupervised learning of disparity maps from stereo images

Notes of this paper which talk about unsupervised learning for stereo match

We treat correspond problem as a transformation from the left to right image.
First, we start with from the assumption that horizontal disparity is caused by horizontal shift. This could be solved by xxx lie group using Fourier theory.

In a second step, we use the general Lie group framework to allow for more general transformations.