The growing number of image acquisition and storage systems in the digital world demand for the new retrieval methods. Most of the existing retrieval methods use textual information, which has been mainly entered manually for every image in the image collection. In order to access the images of interest, user gives textual input against which images are retrieved from the image collection. Sometimes, this results in garbage retrieval due the human involvement in the image annotation process. So more efficient image retrieval mechanism is needed. To overcome the issue, other approach which is generally considered is content-based image retrieval (CBIR). CBIR depends on the automatically extracted features for every image in the image collection as well as their storage and comparison upon a query. Therefore, feature extraction technique and their storage space are important aspects of CBIR. In this paper, we design and develop agent-based CBIR system for image retrieval and suggest the best feature extraction technique in terms of less storage space and more accurate search results. Although the proposed image retrieval technique can be used for any type image collection, our work focuses on the medical images.