Abstract
In this paper, we investigated six distinct partitions of wavelet packets selected from a complete decomposition and measured their performance in terms of sensitivity and selectivity for the classification of twenty-five natural textures. Both energy and entropy metrics were computed for each wavelet packet node and included in distinct features for representation. Each subset of wavelet packet nodes reflected a specific scale and orientation sensitivity. Wavelet packet representations for twenty-five natural textures were classified without error by a simple twolayer network classifier. Texture signatures computed from wavelet packet energy were more reliable than entropy signatures