本文選題：α穩(wěn)定分布　切入點(diǎn)：局部最優(yōu)檢測(cè)　出處：《曲阜師范大學(xué)》2014年碩士論文

【摘要】：網(wǎng)絡(luò)的開(kāi)放性和資源共享給網(wǎng)絡(luò)信息安全帶來(lái)了極大的隱患（如侵權(quán)、篡改等等）。因此，多媒體作品的版權(quán)保護(hù)問(wèn)題亟待解決。近年來(lái)，一種用于知識(shí)產(chǎn)權(quán)保護(hù)的新方法越來(lái)越流行，即在多媒體信息中嵌入數(shù)字水印。數(shù)字水印技術(shù)是通過(guò)某種算法在多媒體數(shù)據(jù)中嵌入特定的信息，具有三大顯著特征：不可見(jiàn)性、魯棒性和安全性。目前絕大多數(shù)檢測(cè)算法采用了線(xiàn)性相關(guān)的方法，由信號(hào)檢測(cè)的基本理論可知，基于線(xiàn)性相關(guān)的水印檢測(cè)方法只有在水印載體服從高斯分布時(shí)才是最優(yōu)的。研究結(jié)果表明，在數(shù)字圖像的時(shí)/空域或者變換域，以高斯分布來(lái)對(duì)載體圖像進(jìn)行統(tǒng)計(jì)建模是不合適的。因此，，從水印檢測(cè)的角度來(lái)看，線(xiàn)性相關(guān)水印檢測(cè)方法沒(méi)有考慮到載體圖像的實(shí)際統(tǒng)計(jì)分布特性，其優(yōu)化條件不復(fù)存在，檢測(cè)性能也隨之嚴(yán)重退化。 對(duì)于離散余弦變換域（DCT）圖像水印，數(shù)據(jù)呈重尾分布，相關(guān)檢測(cè)顯然不是最優(yōu)的檢測(cè)方案。不可感知性是數(shù)字水印的一個(gè)基本特征，這就決定了水印信號(hào)的檢測(cè)是一個(gè)弱信號(hào)的檢測(cè)問(wèn)題。研究表明非線(xiàn)性接收器特別適合呈重尾分布的噪音弱信號(hào)的檢測(cè)，是局部最優(yōu)檢測(cè)方案。這促使了局部最優(yōu)柯西非線(xiàn)性檢測(cè)和零記憶非線(xiàn)性檢測(cè)在DCT變換域圖像中的應(yīng)用。本文主要工作如下： 首先，介紹DCT圖像水印相關(guān)理論基礎(chǔ)知識(shí)，包括DCT域圖像水印的生成、嵌入和檢測(cè)結(jié)構(gòu)，為下面研究DCT域圖像水印的檢測(cè)算法做鋪墊。 其次，對(duì)幾種常見(jiàn)的DCT系數(shù)模型進(jìn)行了分析，并提出一種新的數(shù)據(jù)建模方法，即均衡穩(wěn)定簇模型，并呈現(xiàn)使用穩(wěn)定分布的方法對(duì)圖像的DCT系數(shù)進(jìn)行建模的結(jié)果。 然后，提出了兩種局部最優(yōu)非線(xiàn)性檢測(cè)算法，即局部最優(yōu)柯西非線(xiàn)性檢測(cè)和零記憶非線(xiàn)性檢測(cè)。通過(guò)計(jì)算似然比，對(duì)檢測(cè)性能進(jìn)行理論分析。 最后，對(duì)提出的兩種非線(xiàn)性檢測(cè)算法進(jìn)行量化攻擊，通過(guò)計(jì)算似然比，對(duì)檢測(cè)性能進(jìn)行理論分析，并通過(guò)仿真實(shí)驗(yàn)繪制接收機(jī)工作特性曲線(xiàn)(ROC)，對(duì)檢測(cè)器的性能及魯棒性進(jìn)行比較分析。
[Abstract]:The openness and resource sharing of network bring great hidden trouble (such as infringement, tampering, etc.) to network information security.Therefore, the copyright protection of multimedia works needs to be solved.In recent years, a new method for intellectual property protection is becoming more and more popular, that is, embedding digital watermarking into multimedia information.Digital watermarking technology is to embed specific information in multimedia data through some algorithm, which has three characteristics: invisibility, robustness and security.At present, most detection algorithms adopt linear correlation method. From the basic theory of signal detection, the linear correlation based watermarking detection method is optimal only when the watermark carrier is distributed from Gao Si.The results show that it is not appropriate to use Gao Si distribution to model the carrier image in the time / spatial domain or transform domain of the digital image.Therefore, from the point of view of watermark detection, linear correlation watermarking detection method does not take into account the actual statistical distribution of the carrier image, its optimization conditions no longer exist, and the detection performance is seriously degraded.For DCT image watermarking in discrete cosine transform domain, the data is heavy-tailed, and correlation detection is obviously not the optimal detection scheme.Imperceptibility is a basic feature of digital watermarking, which determines that the detection of watermark signal is a weak signal detection problem.It is shown that the nonlinear receiver is especially suitable for the detection of noise weak signals with heavy-tailed distribution and is a local optimal detection scheme.This results in the application of local optimal Cauchy nonlinear detection and zero memory nonlinear detection in DCT transform domain images.The main work of this paper is as follows:First of all, the basic theory of DCT image watermarking is introduced, including the generation, embedding and detection structure of DCT domain image watermarking, which pave the way for the following research of DCT domain image watermarking detection algorithm.Secondly, several common DCT coefficient models are analyzed, and a new data modeling method, equilibrium stable cluster model, is proposed, and the results of modeling the DCT coefficients of images by using the stable distribution method are presented.Then, two local optimal nonlinear detection algorithms are proposed, that is, local optimal Cauchy nonlinear detection and zero-memory nonlinear detection.The detection performance is theoretically analyzed by calculating likelihood ratio.Finally, the proposed two nonlinear detection algorithms are quantitatively attacked, and the detection performance is theoretically analyzed by calculating likelihood ratio.The performance and robustness of the detector are compared and analyzed by drawing the operating characteristic curve of the receiver through simulation experiments.
【學(xué)位授予單位】：曲阜師范大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2014
【分類(lèi)號(hào)】：TP309.7

【參考文獻(xiàn)】

相關(guān)期刊論文 前4條

1 孫中偉,朱巖,馮登國(guó);DCT域圖像水印的局部?jī)?yōu)化檢測(cè)性能研究[J];電子學(xué)報(bào);2005年05期

2 毛家發(fā);林家駿;戴蒙;;圖像DCT系數(shù)的α穩(wěn)態(tài)簇分布模型[J];華東理工大學(xué)學(xué)報(bào)(自然科學(xué)版);2009年03期

3 伯曉晨,沈林成,常文森,牛軼峰;DCT變換域盲圖像水印的自適應(yīng)檢測(cè)[J];計(jì)算機(jī)研究與發(fā)展;2002年04期

4 朱春江,黃樹(shù)成;擴(kuò)頻技術(shù)在數(shù)字水印技術(shù)中的應(yīng)用[J];連云港職業(yè)技術(shù)學(xué)院學(xué)報(bào)(綜合版);2000年04期

本文編號(hào)：1713396