Effective Multi-Resolution Transform Identification for Characterization and Classification of Texture Groups

Texture classification is important in applications of computer image analysis for characterization or classification of images based on local spatial variations of intensity or color. Texture can be defined as consisting of mutually related elements. This paper proposes an experimental approach for identification of suitable multi-resolution transform for characterization and classification of different texture groups based on statistical and co-occurrence features derived from multi-resolution transformed sub bands. The statistical and co-occurrence feature sets are extracted for various multi-resolution transforms such as Discrete Wavelet Transform (DWT), Stationary Wavelet Transform (SWT), Double Density Wavelet Transform (DDWT) and Dual Tree Complex Wavelet Transform (DTCWT) and then, the transform that maximizes the texture classification performance for the particular texture group is identified.