A nature texture classification method based on fractal dimension using self-similar texture characterization is presented. For this purpose, eight fractal dimensin (FD) features are used based on the original image, the high gray level image, the low gray level image, four directional (0°, 45°, 90°, 135°) gradient image and the multi-fractal dimension of order two. The fractal dimension is estimated with the modified box-counting method. The texture classification is completed with back propagation (BP) neural network. The results are compared with other techniques. The Wiener filter is used to improve the performance of this method.