Image Retrieval with Relevance Feedback using SVM Active Learning

Abstract

In content-based image retrieval, relevant feedback is studied extensively to narrow the gap between low-level image feature and high-level semantic concept. In general, relevance feedback aims to improve the retrieval performance by learning with user's judgements on the retrieval results. Despite widespread interest, but feedback related technologies are often faced with a few limitations. One of the most obvious limitations is often requiring the user to repeat a number of steps before obtaining the improved search results. This makes the process inefficient and tedious search for the online applications. In this paper, a effective feedback related scheme for content-based image retrieval is proposed. First, a decision boundary is learned via Support Vector Machine to filter the images in the database. Then, a ranking function for selecting the most informative samples will be calculated by defining a novel criterion that considers both the scores of Support Vector Machine function and similarity metric between the "ideal query" and the images in the database. The experimental results on standard datasets have showed the effectiveness of the proposed method.