Scalable Integrated Region-Based Image Retrieval Using IRM and Statistical Clustering.

Statistical clustering is critical in designing scalable image retrieval systems. This paper presents a scalable algorithm for indexing and retrieving images based on region segmentation. The method uses statistical clustering on region features and IRM (Integrated Region Matching), a measure developed to evaluate overall similarity between images that incorporates properties of all the regions in the images by a region-matching scheme. Compared with retrieval based on individual regions, this overall similarity approach: (1) reduces the influence of inaccurate segmentation; (2) helps to clarify the semantics of a particular region; and (3) enables a simple querying interface for region-based image retrieval systems. The algorithm has been implemented as part of an experimental SIMPLIcity image retrieval system and tested on large-scale image databases of both general-purpose images and pathology slides. Experiments have demonstrated that this technique maintains the accuracy and robustness of the original system while reducing the matching time significantly. (Contains 41 references.) (Author)