A new effective method for image classification is presented in the paper based on feature weighting and support vector machine.First of all,each dimension of the feature is set different weight according to its stability judged by the degree of its importance.And then,the images are classified based on the weighted feature and support vector machine.Finally,different color and texture features are adopted to verify the classification accuracy in the case of feature weighted and unweighted.Experimental results show that the proposed method is effective to improve the accuracy of image classification.