Content-based image retrieval using wavelet-based feature extraction method

This paper proposes a new image index and retrieval algorithm conducted in SPIHT compression wavelet domain, where image features are extracted from the modified lowest frequency subbands. Hybrid low level image features including color and texture are applied in this paper. Sharp parts or edges or the image are kept in color feature description by combining selected wavelet salient points. The resulting indexing technology is proven to improve retrieval results in terms or retrieval accuracy, computational cost and storage space or feature vectors in general image library.