Visual surface reconstruction and boundary detection using stochastic models

The specification of regularization parameters is one of the difficult problems in using the weak membrane models for visual surface reconstruction and boundary detection. The gradient limit effect is a fundamental limitation of these models. In this study, we reduce the gradient limit effect by fusing the intensity and the range image of the same scene utilizing the Markov Random Field (MRF) models. In order to improve the reconstruction we propose an extended weak membrane model that exhibits more complex interactions of the line process as well as the intensity and the depth processes. Consequently, the feasible regularization parameter space becomes larger, resulting in a considerably independent reconstruction on the model parameters. The performance of the introduced model has been quantitatively evaluated by using a Kolmogorov-Smirnov (KS) difference measure.