A Low Complexity Wavelet-Packets Based Image Signature for Indexing and Retrieval

Image signature is a signal that is significantly shorter and simpler than the image signal and can be used to efficiently identify the original image. Generating an image signature that accurately represents an image can only be done by means of extracting specific features of the image and express it in an efficient way. In this paper, a new wavelet-packets-based (WP) image signature that can be used for image indexing and fast retrieval is proposed. This signature is generated via encoding the position of specific points of the signal that represents some statistics of the wavelet packets coefficients in the various frequency bands. Computer simulation results using nearly 7500 image signatures show that the targeted image signature was identified with a very high accuracy in a very short period of time. Also, it showed that the proposed signature is robust against geometric transformation