Stereo matching using colour consistency and hierarchical stochastic diffusion

A stereo matching algorithm is proposed using colour consistency and hierarchical stochastic diffusion. The colour consistency is introduced to the prior Markov random field (MRF) model to consider both the smoothness and the discontinuity of the disparity field. The energy function derived from the MRF model is minimised in the hierarchical stochastic diffusion, which first propagates the probability distribution of the reliable disparity field into the neighbourhood. Experimental results show good disparity maps compared with other propagation based methods.