摘要

Characterized by its easy duplication, storage and dissemination, digital contents are increasingly involved in copyright protection. This paper, based on the matrix singular value decomposition in contourlet domain, suggests an adaptive watermarking scheme that embodies four major algorithms including watermark preprocessing, HVS masking computation, watermark embedding and extracting. Firstly, the HVS masking computation and the image preprocessing are conducted before contourlet transform is performed on the host image to extract lowpass images and one directional subband image. Then, watermarks are embedded into the above resulted images and the extracting is an inversion of the embedding. The experiment has demonstrated the scheme enjoys good imperceptibility, security and robustness in that the watermarked image is capable of extracting watermark via the lowpass image or the directional bandpass image even when attacked by compression, noising, cutting, translation and others.

全文