This paper describes ExSight, a prototype system for content-based image retrieval that will provide image retrieval facilities based on the indexing of component objects. We present a database centric approach to image retrieval and other techniques necessary for successfully implementing ExSight. The essential point of this approach is automatic image data analysis, emphasizing automatic object extraction that implies redundancy. The database module of ExSight coordinates multiple space indices in order to obtain an overall ranking based on several different features. The experimental results reveal that object-based contents retrieval achieves a higher level of retrieval correctness than color-region based retrieval and the implemented multidimensional data access engine achieves real-time response.