Abstract: Applying support vector machine (SVM) to the contentbased image retrieval, the novel method of integrated texture recognition based on SVM and Gabor wavelet transform was proposed. Texture feature extracted by using multichannels Gabor wavelet filters was denoted by the coefficients of Gabor wavelet transform and its standard variance. In addition, the normalized feature vectorswere used to train and identify by support vector machine. Finally, experiments and comparisions with other methods using Brodatz texture database indicate that the presented method provide superior texture recognition accuracy and the best generalization ability over other methods under the condition of limited training samples.

Key words: support vector machine, feature extraction, texture recognition, contentbased image retrieval , Gabor wavelet transform