MIR: An approach to robust clustering - Application to range image segmentation

This paper describes an unsupervised region merging technique based on a no vel robust statistical test. The merging decision is derived from the mutua l inlier ratio (MIR) of adjacent regions. This ratio is computed using robu st regression techniques and a novel method to estimate the robust scale of the Gaussian distribution. A discrimination value to recognize identical G aussian distributions with the MIR is derived theoretically as a function o f the sizes of the compared sets. The presented method to test distribution s is compared with the established Kolmogorov-Smirnov test and implemented into a segmentation algorithm for planar range images. The iterative region growing technique is evaluated using an established framework for range im age segmentation comparison involving 60 real range images. The evaluation incorporates a comparison with four state-of-the-art algorithms and gives a n experimental demonstration of the need for robust methods capable of hand ling noisy data in real applications.