Relevance Feedback Content-Based Image Retrieval Using Query Distribution Estimation Based on Maximum Entropy Principle

In the last few years, we have seen an upsurge of interest in content-based image retrieval (CBIR){ the selection of images from a collection via features extracted from images themselves. Typically the nearest-neighbor rule is used to retrieve images from a query image. However, the underlying query distribution may not be isotropic in nature. Hence, a more sophisticated estimation for the query distribution is required. We propose a novel relevance feedback framework for image retrieval which contains two stages: (1) to estimate the query distribution based on relevance feedback information and (2) to generate a set of inquiries for relevance selection based on the MaximumEntropy Principle. We demonstrate these two stages in detail. Moreover, experiments have been performed on a trademark image database. The results show our proposed framework is e ective in image retrieval with a few relevant samples.