Image classification based on texture features plays an important role in content-based image retrieval. A novel method for rotation invariant image classification is proposed based on MPEG-7 Homogeneous Texture Descriptor (HTD). To compute HTD, Gabor transform is first performed by filtering image using a bank of orientation and scale selective band-pass filters called Gabor wavelets. For constructing the feature vector, the mean energy and standard deviation are calculated separately on each filtered and original image. Then the summations of energy in different direction are calculated and the direction with the maximum energy is chosen as the dominant orientation, which is used to shift the feature vector circularly in order to keep rotation invariant. The classification is done by SVM (Support Vector Machine). The method is tested on Brodatz and UIUCTex datasets. The experiments demonstrate the method is effective and efficient for rotation invariant texture classification.