Hamed Aghapanah - Tarbiat Modares University
Hassan Ghassemian

Texture is an important characteristic for the analysis of main type of multi texture images. In This paper, we present a statistical method to find those spatial (or other)relations that best detention the structure of textures when some feature extraction method is used. These selectionsshould accordingly be well suited for discriminations that arestructurally based. Adaptive power spectral density (PSD) is a good method to classify the separate textures with verity PSDbut in dermoscopy images' classification; we must try to select best features. Because some features are correlated togetherand doesn’t have informative value. So we purpose a method to find best features from some methods. Experimental results show better performance in this method with acceptable wasted time. costume acceptable time when compared withprevious methods. Experimental results show the accuracy and validity of the classification increased up to 83.66

Fisher discrimination, PSD, Ripplet transform, Gabor filter bank, CBIR