Image Retrieval Algorithm Based on Feature Fusion and Bidirectional Image Matching

摘要

According to the low accuracy of single feature-based image retrieval algorithm, the paper proposes a strategy which fuses the SIFT and the color feature, to obtain a kind of relatively simple and accurate retrieval approach. To accelerate the retrieval process and improve the efficiency, the paper adopts the image feature-based hierarchical retrieval method. The RGB col-or feature is the first layers of retrieval feature, presented by the color histogram. SIFT is the second layer of retrieval feature. Meanwhile, for the SIFT retrieval process, by using the bidirectional matching strategy, the improvement of the image retrieval precision is expected. Experiment results show that both the retrieval rate and accuracy are efficiently improved.