An Image Retrieval Method Based on Collaborative Filtering
Relevance feedback plays an important role in image retrieval. As a short-term learning strategy, it learns from the user’s relevance evaluation on the current retrieval’s output result to improve the retrieval performance. Nowadays using long-term learning strategy to improve image retrieval attracts more and more attention. In this paper, we present a composite image retrieval approach using both of them to improve image retrieval. Our approach is based on on-line analysis of feedback sequence log, the archive of the user’s feedback evaluation data sequence created in the past. For long-term learning, Collaborative Filtering is adopted to predict the semantic correlations between images. During CF process, we make use of Edit Distance to evaluate the similarity between the feedback sequence records. Experiments over 11,000 images demonstrate that our method achieves significant improvement in retrieval effectiveness compared with conventional method.
Unable to display preview. Download preview PDF.