Similarity Measure of the Visual Features Using the Constrained Hierarchical Clustering for Content Based Image Retrieval

Abstract

In this paper, we present a methodology on how to measure the visual similarity between a query image and hierarchically represented image databases for content based image retrieval. The images in database are hierarchically summarized and classified by recovered extrinsic camera parameters as well as constrained agglomerative clustering methods. The constrained agglomerative hierarchical image clustering method whose strategy is to extract a multi-level partitioning and grouping of multiple images is used for balancing the hierarchical trees and summarization. The visual codebooks which are hierarchically quantized in the clusters are used to calculate the similarity measure with a query image's visual features. Our proposed visual similarity measure and summarization of image data provide a very efficient way for searching and retrieving the images that have similar visual contents and geometrical location.