The purpose of this research is to develop a new visual information analysis, representation, and retrieval framework for automatic discovery of salient objects of user's interest in large-scale image databases. In particular, this dissertation describes a content-based image retrieval framework which supports multiple-object retrieval. The proposed research, unlike most existing works that are designed for single object retrieval or adopt heuristic multiple object matching scheme, aims at contributing to this field through the development of an image retrieval system that enables effective and efficient multiple-object retrieval and automatic discovery of the objects of interest through users' relevance feedback. The key to achieving the above goal is a new systematic and hierarchical image representation, and a related learning and retrieval framework, the combination of which makes it possible for a machine to interpret an image in terms of its containing regions and their relationships. In this dissertation, an efficient and accurate hierarchical image segmentation algorithm based on multi-resolution analysis is developed to alleviate the over- and/or under-segmentation problems through the preservation of associative relationships between image regions in a hierarchical region-tree. This algorithm is designed in a way to simultaneously produce image segmentation results and hierarchical region-tree representations, which are typically obtained through two separate processes in existing approaches, so as to reduce the time complexity. With hierarchical region-tree representations, the relevance of an image to the query image is thus measured according to the sub-tree comparison. As a full comparison of all sub-trees is unlikely to be feasible, an efficient strategy for selecting and comparing proper sub-trees is designed and developed. Another key contribution of this research is the seamless integration of users' relevance feedback (RF) with the proposed multiple object retrieval system, which allows automatic discovery of the objects of users' interest and is expected to improve the retrieval accuracy through feedback-retrieval loops. While there is a clear indication of needs for such interactive learning capabilities, we believe this is the first systematic attempt to formulate a comprehensive, intelligent, and interactive framework for multiple object retrieval in image databases that takes full advantage of a hierarchical region-tree representation. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]