How to improve the accuracy of contentbased image retrieval is a difficult problem, while its key point is feature extraction method. Aiming at the deficiency of single feature image retrieval, we propose an effective image retrieval method based on multifeatures of images. We firstly use color and edge directivity descriptor (CEDD) to extract color and texture features of the images, then use pyramid histogram of oriented gradients (PHOG) to extract shape features, finally we adopt the feature serial correlation method to integrate the three image features into retrieval. The experiment using Corel image sets shows that the precision and the AvgP value of our method both perform better than the two original algorithms.