In this paper, we have investigated an approach based on log-polar transform and support vector machine (SVM) for texture classification. Firstly, we convert rotation into translation, which reduced the effect of rotation and scale changes using the characteristics of logarithm Polar transformation. Then we extract texture feature after eliminating translation using smooth discrete wavelet transform (SWT) with translation invariant. Finally, support vector machines(SVM) are adopted to the texture classification. The experiment results show the proposed approach can get improved results for texture classification.