Skip to Main Content
Due to the recent improvements in digital photography and storage capacity, storing large amounts of images has been made possible, and efficient means to retrieve images matching a user's query are needed. Content-based Image Retrieval (CBIR) systems automatically extract image contents based on image features, i.e. color, texture, and shape. Relevance feedback methods are applied to CBIR to integrate users' perceptions and reduce the gap between high-level image semantics and low-level image features. In the past 30 years, relevance feedback (RF) has been an effective query modification approach to improving the performance of information retrieval (IR) by interactively asking a user whether a set of documents are relevant or not to a given query concept. This paper aims at developing a scheme for intelligent image retrieval using machine learning technique and the information gathered from the user's feedback. This helps the system on the following rounds of the retrieval process to better approximate the present need of the user. We have shown that a powerful relevance feedback mechanism can be implemented by using reinforcement learning algorithms. The user thus does not need to explicitly specify weights for relationship between images and concepts, because the weights are formed implicitly by the system. The proposed relevance feedback technique is described, analyzed qualitatively, and visualized in the paper. Also, its performance is compared with a reference method. Experimental results demonstrate that our proposed technique is promising.