水印技术中相似度公式的一种改进算法

An Improved Algorithm of Similarity Formula in Digital Watermarking

目前国内外在评价数字水印技术中水印的相似程度主要采用的是归一化相关值(Normalized Correlation值),尽管常见的几种归一化相关值公式可以客观的表示原始水印与提取水印的相似程度,但如果要从人类视觉的主观角度出发,这些传统的归一化相关值公式在表示提取出的水印和原始水印反面相似时会存在很大不足。为此,提出了一种基于图像像素均值的数字水印相似程度的评价公式,文章称之为均值相关值。使用均值相关值公式可以有效的继承传统的归一化相关值,并克服了这些传统评价相似度的不足。通过仿真实验证明文章构建的均值相关值公式具有很好的科学性、合理性和有效性。

Presently normalized correlation value has been used in digital watermarking evaluation system at home and abroad.Although traditional normalized correlation formula can shows the similarity between original watermark and extracted watermark,it also exists some shortcomings while showing counter-similarity between original watermark and extracted watermark when we consider the HVS system.Accordingly this paper proposes one digital watermarking evaluation formula about watermark similarity value based on average value of image＇s pixels,which we call average correlation value.Using average correlation value formula can inherit traditional normalized correlation formula,meanwhile,which can overcome some disadvantages of traditional similarity value.The experimental results demonstrate this proposed average correlation value formula has good science,rationality and efficiency.