ROBUST ESTIMATION FOR RANGE IMAGE SEGMENTATION AND RECONSTRUCTION

This correspondence presents a segmentation and fitting method using a new robust estimation technique. We present a robust estimation metho d with high breakdown point which can tolerate more than 80% of outlie rs. The method randomly samples appropriate range image points in the current processing region and solves equations determined by these poi nts for parameters of selected primitive type. From K samples, we choo se one set of sample points that determines a best-fit equation for th e largest homogeneous surface patch in the region. This choice is made by measuring a RESidual Consensus (RESC), using a compressed histogra m method which is effective at various noise levels. After we get the best-fit surface parameters, the surface patch can be segmented from t he region and the process is repeated until no pixel left. The method segments the range image into planar and quadratic surfaces. The RESC method is a substantial improvement over the least median squares meth od by using histogram approach to inferring residual consensus. A gene tic algorithm is also incorporated to accelerate the random search.