This paper present a novel image retrieval framework depend on a knowledge base system. The proposed framework composed of three techniques namely; feature extraction technique, image clustering technique and rule base technique respectively. The first one extracts the features from the images database using wavelet transform (WT). Therefoe, each image can be characterized by a few coefficients which are independent on the image resolution. The second technique clusters the images database according to four phases called; features' division, mixture of Gaussian model, relevance feedback and pruning model. Consequently, at the end of the clustering technique each image in the database has been belonged to a specific cluster (supervised learning). The third technique makes use a genetic algorithm (GA) as a tool for extracting a set of rules from these images and the corresponding classes. The fitness function of the genetic algorithm is evaluated according to the support and the confidence measures. Finally, the results presented in this paper demonstrate that the proposed model provides accurate retrieval results and achieve improvement in performance with significantly less computation time compared with other models.