TEXTURAL ANALYSIS OF RANGE IMAGES

This paper addresses the problem of texture discrimination in range images. The range data is transformed to a common coordinate system, and then resampled to generate a regular grid of points in order to eliminate the effect of different sampling rates. Eight statistical textural features based on co-occurrence matrices are computed on the resampled data, and are used to discriminate between two classes of natural surfaces consisting of pebbles of different sizes lying on a plane. Seven of the eight textural features are found to be useful in discriminating between these two classes. The experiments also confirm the importance of resampling the data before computing the textural features.