Substantial accumulations of computerized photos are consistently made. Various collections of digital photos like in the area of technology, private organization, etc. are structured because of the result of digitizing existing accumulations of simple pictures, charts, outlines, canvases, and archives. For the most part, the main strategy for investigating these accumulations was by perusing or indexing of index words, Digital photo databases begin the best approach to searching based on content. In this paper, we have shown a strategy that has no past information about the picture inside the database, yet retrieval is carried out considering the content data of the pictures prone to be called as content based on image retrieval. Here we are attempting to enhance the image retrieval framework for more exactness and efficiency by utilizing Radial base Function neural system. This arrangement with multilayer feed forward system recognition. By utilizing these procedures we can get into know the exact relevant image gave by the client according to the query. The Scale Invariant Feature Transform (SIFT) is a standout amongst the most nearby feature detector and descriptors which is utilized as a part of the vast majority of the vision programming. In this paper in regards to CBIR framework we can use SIFT algorithm to concentrate the nearby feature of the pictures. The retrieval method has been implemented on two CBIR systems, one using multiple features and the other using only RBG feature. The results obtained are positive and we have obtained higher precision (81.93%) using multiple features than that of using only RGB (41.18%). Experiment demonstrates that the proposed system get better results of retrieval system.