This paper presents a Curvelet Transform -based texture analysis method for classification of non-melanoma skin lesion classification. We are applying tree-structured on wavelet transform and subband analysis on Curvelet Transform on gray scale image analysis on wavelet and curvelet coefficients. In our proposed method Feature extraction and a 8 subband stages on  feature selection method, based on entropy and correlation, were applied to a train set of  images. The resultant feature subsets were then fed into neural network  classifiers on 3 stages of Normal;Disease Effected 30-50% and  Disease Effected above 50% and comparision measurement analysis on DWT /DCT Using Probabiliste Neural Network Classifier  based accuracy improvement.