A Novel Mesh adaptation Technique for the computation of Stereo Depth Using Texture less Simple Stereo Pairs

Abstract

A novel mesh adaptive strategy for a stereo vision depth for texture-less image pairs is proposed in this paper which is based on a posteriori error estimators. A Variational approach is proposed to solve one directional disparity map problem using TV regularization model. An adaptive P1 finite Element method is used as a simulation device for the numerical solution of PDEs based model which is obtained from the minimization of the chosen energy model. This mesh adaptation procedure is performed by a posteriori error indicators in appropriate norm. Our locally adaptive methodology allows us the local and automatic choice of smoothing parameters.