TEXTURE CLASSIFICATION BY WAVELET PACKET SIGNATURES

This correspondence introduces a new approach to characterize textures at multiple scales. The performance of wavelet packet spaces are meas ured in terms of sensitivity and selectivity for the classification of twenty-five natural textures. Both energy and entropy metrics were co mputed for each wavelet packet and incorporated into distinct scale sp ace representations, where each wavelet packet (channel) reflected a s pecific scale and orientation sensitivity. Wavelet packet representati ons for twenty-five natural textures were classified without error by a simple two-layer network classifier. An analyzing function of large regularity (D20) was shown to be slightly more efficient in representa tion and discrimination than a similar function with fewer vanishing m oments (D6). In addition, energy representations computed from the sta ndard wavelet decomposition alone (17 features) provided classificatio n without error for the twenty-five textures included in our study. Th e reliability exhibited by texture signatures based on wavelet packets analysis suggest that the multiresolution properties of such transfor ms are beneficial for accomplishing segmentation, classification and s ubtle discrimination of texture.