为了实现大容量、低感知度的数字图像隐写,根据人眼视觉系统对图像边缘变化不敏感的特点,提出一种结合图像多尺度边缘检测和最小误差替换的自适应算法.简单介绍了常用的图像边缘检测算法以及Canny算子,而后通过利用Canny算子检测图像边缘像素.再按一定比例将秘密信息分别嵌入到边缘像素和普通像素中,并对比选用不同嵌入比例时算法的隐写容量和峰值信噪比差异,选定合适的嵌入比例.最后采用最小误差替换方法修正图像峰值信噪比,降低图像失真度,并对比算法针对不同载体时的嵌入容量和峰值信噪比.实验结果表明:以图片“Baboon”作为载体按4∶1嵌入比例隐写秘密信息时,本方案嵌入容量为525 612 bit,载秘图像峰值信噪比为413 dB.该算法不仅隐写容量十分可观,而且失真度小,计算简单,易于实现.

In order to realize a image steganography with a high hiding capacity and undetectability,based on the fact that the changes in image edges are insensitive to human eyes,a novel adaptive algorithm employing multiple edge detection and Minimum Error Replacement (MER) is presented.By Canny operator,large numbers of edge pixels are picked out and the secret information can be embeded into the edge pixels and common pixels to a certain proportion.To compensate for the resulting decrease in the Peak Signal to Noise Ratio (PSNR),MER method is used.Experimental results indicate that when the Baboon is chosen as cover image and the proportion of embedding the secret information is 4∶1,the hiding capacity is 525 612 bit and PSNR is 413 dB.This algorithm not only has a high hiding capacity and good visual qualities,but also easy to carry out.