In this paper, a novel texture classification method using selected and combined features from wavelet frame and steerable pyramid decompositions has been proposed. Firstly, wavelet frame and steerable pyramid decompositions are used to extract complementary features from texture regions. Then the number of features is reduced by selection using maximal information compression index. Finally the reduced features are combined and forwarded to SVM classifiers. The experimental results show that the proposed method used selected and fused features can achieve good classification accuracy and have low computational complexity.