Likelihood and Confidence of Terrain Profiles in the Symmetric Dynamic Programming Stereo

The paper considers probabilistic regularisation of partial occlusions in the intensity-based symmetric dynamic programming stereo (SDPS). The regularised SDPS relates the total similarity of corresponding pixels to a cumulative log-likelihood ratio for a profile that yields the correspondences. Transition probability models are investigated in detail in order to explicitly relate the likelihood ratios to reconstruction parameters. Possibilities of defining confidence of the reconstruction in terms of local distributions of the cumulative log-likelihood ratios along the GPV are discussed. 1 Center for Image Technology and Robotics Tamaki Campus, The University of Auckland, Auckland, New Zealand. [email protected] You are granted permission for the non-commercial reproduction, distribution, display, and performance of this technical report in any format, BUT this permission is only for a period of 45 (forty-five) days from the most recent time that you verified that this technical report is still available from the CITR Tamaki web site under terms that include this permission. All other rights are reserved by the author(s). Likelihood and Con dence of Terrain Pro les in the Symmetric Dynamic Programming Stereo Georgy Gimel'farb and Tao Wu CITR, Department of Computer Science Tamaki Campus, University of Auckland Private Bag 92019, Auckland 1 [email protected]