Segmentation based on fusion of range and intensity images using robust trimmed methods

This paper proposes a segmentation algorithm based on fusion of range and i ntensity images using robust trimmed methods. Based on the Bayesian theory, a priori knowledge is represented using the Markov random field (MRF). A m aximum a posteriori (MAP) estimator is constructed using the edge features extracted from both range and intensity images. Objects are represented by a number of local planar surfaces in range images, and the parametric space for surface representation is constructed with the surface parameters esti mated pixel-by-pixel based on the least trimmed squares (LTS) method. Where as in intensity images, the alpha -trimmed variance is adopted as the featu re for edge extraction. A final edge map is obtained by the MAP estimator t hat is constructed using the likelihood functions based on the edge informa tion obtained from range and intensity images. Finally, an image is segment ed using the fused edge map. Computer simulation results show that our new segmentation algorithm effectively segments test images, independent of sha dow, noise, and lighting environment. (C) 2001 Pattern Recognition Society. Published by Elsevier Science Ltd. All rights reserved.