In recent years, digital watermarking technique has become an active research topic. People find that it is inconvenient to extract watermarks with the original image, so public watermarking algorithm has been discussed widely. In this thesis, we study how public watermarking solves application problems, and propose some algorithms to solve the difficulties in public watermarking such as scaling attack, VQ attack etc. The main contributions of this thesis include the following : ① We have proposed a public robust algorithm based on spread spectrum. The basic principle of the algorithm is derived from the relationship between the DCT coefficients of images before and after scaling. Watermark extraction is independent of the original image or any other conditions. Semantically meaningful watermarks can be embedded into a host image. This approach can resist not only common signal processing attacks such as compression and filtering etc, but also scaling attacks as well. ② We have proposed a self-recovery algorithm based on regions of interest. This algorithm divides a host image into two parts: regions of interest and regions of background. We code the regions of interest based on 4× 4 blocks, and embed these codes into the DCT coefficients of the background by a novel mapping function. Our scheme can authenticate the whole image, and at the same time, if the region of interest is tampered maliciously, the scheme can recover the original content in good quality. ③ We have introduced a double-watermarking algorithm to authenticate in- ages, recover the original image content, as well as to protect copyright. This technique firstly encodes a serial number in a logo, and then inserts this logo into an original image. This algorithm can detect malicious tamper and recover the original image content clearly.
修改评论