Sample Selection Strategies for Relevance Feedback in Region-Based Image Retrieval

摘要

The success of the relevance feedback search paradigm in image retrieval is influenced by the selection strategy employed by the system to choose the images presented to the user for providing feedback. Indeed, this strategy has a strong effect on the transfer of information between the user and the system. Using SVMs, we put forward a new active learning selection strategy that minimizes redundancy between the examples. We focus on region-based image retrieval and we expect our approach to produce better results than existing selection strategies. Experimental evidence in the context of generalist image databases confirms the efectiveness of this selection strategy.