A New Similarity Measurement Based on Distance and Correlation Test for Content-Based Images Retrieval

While extracting feature vectors to represent image content remains an important issue especially those semantic features, how to measure the similarity of these feature vectors and maximize their relevance to visual content becomes an attractive research topic over the recent years. Following this trend, we proposed a new similarity measurement by introducing a correlation test into the conventional distance matching mechanism for content-based image retrieval and illustrate that such proposed scheme achieves performance improvement upon an existing counterpart. And the proposed scheme can be applied to any existing content-based image retrieval algorithm.