Range image segmentation using multiple markov random fields

Abstract

In this thesis two algorithm are proposed. One is a range image segmentation algorithm and the other is a hexagonal edge detection algorithm. The proposed range image segmentation consists of depth smoothing, gradient smoothing, edge detection and surface type labeling step using Markov random field (MRF)s. The problem of each step is defined as a energy minimization problem using (MRF)s, and the minimum energy state is found using iterative method. First, range and its gradient images are smoothed preserving jump and roof edges respectively using line process concept one after another. Then jump and roof edges are extracted, combinated, and refined using penalizing undesirable edge patterns. Finally, curvatures are computed and the surface type labels are refined using winner-takes-all layers in the last stage. The final output is a set of regions with its exact surface type. Several experimental results show the robustness and generality of our approach and the execution speed of the proposed method is faster than that of a typical region merging method for nearly equal quality of segmentation. In addition, since the MRF formulation generates a local algorithm our algorithm can be easily paralleized. This promises practical applications of our method. This thesis also proposed a method of boundary detection in a hexagonal radiative heat sources are used. The effect of the in-plane temperature gradient at the solicification interface is secondary. In the reflectivity-change dominated regime, the solidification interface has cellular morphologies while in the emissivity-change dominated regime, small faceted interface morphologies develop resulting in closely spaced, frequently branching subboundary patterns. The substrate melting and twin boundary formation also occur in the emissivity-change dominated regime. Therefore, ZMR operation must always be performed in the reflectivity-change dominated regime to obtain good quality films. We obtained the best result ...