Texture Classification Based On Nsci And Brushlet

Separable 2D wavelet transform can provide an optimal representation for one-dimensional piecewise smooth signals, but it can't provide an optimal representation in multidimensional condition. The purpose and impetus to develop the Multiscale Geometric Analysis (MGA) systems is to provide an optimal representation for a new multidimensional function.Nonsubsampled Contourlet Transform(NSCT) inherits the properties of multiscale and multidirection of Contourlet transform, it also has shift invariant property and inducts more redundant information and effective direction decomposes. NSCT based method is used to texture classification in the paper. It has proved that this method is effective both in symmetrical and unsymmetrical texture classification via experiments.Brushlet transform has successfully used in image compression, denoising, classification and so on. It maybe not an ideal choice when Brushlet transform is used to texture classification. But feature extraction method based on NSCT and Brushlet transform have much complementarity. Combining the features of NSCT with Brushlet transform, a method based on NSCT and Brushlet is introduced and applied to Brodatz texture classification. It has proved that this method increases the classification accuracy compared with other methods in symmetrical texture classification. And also, for unsymmetrical texture classification, experimental results show that our method has much higher classification accuracy.