127 
RESEARCH ON THE AUTOCORRELATION DETECTION IN DIGITAL 
WATERMARKING TECHNIQUES 
Cheng-song YANG 2 3 , Chang-qing ZHU a ’ b ,Qisheng Wang 3 ’ 0 
a Zhengzhou Institute of Surveying and Mapping, Zhengzhou 450052 - 
ycsdongshang@ 163 .com 
b Ministry of Education Key Laboratory of VGE, Nanjing Normal University, Nanjing 210054 - 
zcq88@263.net 
Purveying and Mapping Support Unit of South Xinjiang,Shule,Xinjiang.844200 - 
qishengw@yahoo.cn 
watermark autocorrelation detection. First, the setting of detection threshold, the calculating of false alarm probability and missing 
the attack strength is established. Third, an improved method of watermark generating is proposed to promote the robustness of 
watermarking algorithms. Finally, the probability of multiple watermark embedding is studied. 
and hiding it in the source data, it makes the copyright 
information as a part of the source data which can not be 
divided easily. The digital watermarking techniques can be Let G = {g(k)},k = 0,1,— 1 be random noise, here 
used to protect copyright. The digital watermarking has been 
applied in many fields such as digital images, video, audio and 
Commission VI, WG IV/1 
KEYWORDS: Digital Watermarking, Autocorrelation Detection, Multiple Watermarking, Robustness 
ABSTRACT: 
Based on the self-characteristic of digital watermarking techniques and the multi-statistic theory, this paper makes a research on the 
detection probability are studied. Second, the relationship between the length of watermark, the watermark embedding strength and 
1. INTRODUCTION 
watermark strength, the reason why choose such watermark is 
The digital watermarking is a new developed technology for that this watermark has better statistic characteristic for 
information security. By embedding copyright information into watermark detection. 
so on (Sun Shenghe, 2004; Yang Yixian, 2006). 
A watermarking system is comprised of watermark generation, 
watermark embedding and watermark extraction/detection. The 
main research of watermarking techniques is on the watermark 
embedding(Li Yuanyuan,2004;Zhong Shangping,2006). 
However, in the process of watermark extraction, the 
watermark generation and watermark autocorrelation detection 
have an important influence on the accuracy and reliability of 
the watermark extraction/detection result. Based on the reasons 
above, under the additive random noise, this paper makes a 
research on the generation of watermark and the 
autocorrelation detection of watermark based on the statistics 
theory, studying the setting of detection threshold, the 
calculating of false alarm probability and missing detection 
probability, the relationship between the length of watermark, 
the watermark embedding strength and the attack strength. 
Finally, the probability of multiple watermark embedding is 
studied. 
w(k) and g(k) is mutually independent, so 
From the central limit theorem, we can get that: 
When n —> +oo 
k=n-1 
So, when the to-be-detected information is attacked by the 
random noise, there are two cases: 
2. BASIC AUTOCORRELATION DETECTION 
THEORY FOR DIGITAL WATERMARKING 
Case 1: Contain watermark 
Let W = {w(k)},k- 0,1,-be the watermark, 
To-be-detected information: 
У = {y(k)} = W + G = {Я * w(k) + g (к)}