Feature Extraction Using Overlay Blocks and Maximal Eigenvalues for Image Retrieval

In order to organize, manage, browse and retrieve image database effectively, we proposed a novel technique for content based image retrieval using overlay blocks and maximal eigenvalues. Because image pixels have strong relationship with each other, the proposed algorithm generates feature vectors that use overlay parts between adjacent image blocks, and generates the binary map using maximal eigenvalues and block truncation technique for image retrieval. It was noticed that the binary map generated using the proposed method could reflect the target object information more efficiently. The experimental results show that the indexes have good ability of characterizing the image contents.