基于OBSA多小波和遗传算法的数字水印技术

针对已有数字水印技术的不足,提出了基于OBSA多小波和遗传算法的公开数字水印技术。根据不同的水印嵌入算法选择适应度函数,通过遗传算法,对不同的水印嵌入算法、不同的宿主图像,自适应地得到OBSA多小波变换的前置滤波器。给出了相应的两个算法,算法1利用“奇偶法”嵌入水印图像,将灰度水印图像分解成位图,宿主图像进行一层多小波分解,每个低频系数嵌入1bit水印信息;若嵌入“1”,使对应的低频系数四舍五入为奇数,嵌入“0”,则使其为偶数,相反亦可。算法2利用“替代法”嵌入水印,宿主图像进行二层多小波分解,将其分解后的中频部分用水印图像替代。算法1提取的水印无失真,抗剪切攻击能力较强;算法2提取的水印无失真,抵抗噪声、压缩等处理的能力较强。该方法可以改善水印系统的鲁棒性,提高水印系统的透明性和不可统计性。

New Approach to Digital Watermarking Based on OBSA Multi-wavelet Transform and Genetic Algorithms

New approach to blind digital watermarking based on the OBSA multiwavelets and genetic algorithms is presented in this paper.OBSA multiwavelets have randomicity as to selection of the pre-filters.A pre-filter Q can be constructed through genetic algorithms according to different fitness function for proposed embedding algorithms.In watermarking algorithm 1,a grey watermark image is decomposed to bit-planes,one level DMWT is realized by using the pre-filter Q and OBSA multiwavelet for host images.The watermark bit-planes are embedded in low frequency parts."0" or "1" of the watermark bit-planes is represented by even or odd number respectively,and vice versa.This method can keep good perceptual transparency,and make no distortion of the extracted watermark.The replacement method is used in watermarking algorithm 2,two level DMWT is realized by using the pre-filter Q and OBSA multiwavelet for host images,a grey watermark image is embedded by replacing the selected band.The experiment results show that this proposed approach has the desired properties such as invisibility,reliable detection and robustness against typical operation.