基于奇異值分解的數(shù)字圖像水印算法研究

電子通信論文編號:TX067        論文字?jǐn)?shù):25226,頁數(shù):54

目 錄

摘 要   I
ABSTRACT   II
第一章  概述   1
1.1  數(shù)字圖像水印的相關(guān)知識   1
1.1.1  數(shù)字圖像水印  1
1.1.2  數(shù)字圖像水印技術(shù)   1
1.1.3  數(shù)字圖像水印技術(shù)的分類   2
1.2  基于SVD的水印算法研究現(xiàn)狀   4
1.3  本論文章節(jié)安排   6
第二章  算法相關(guān)技術(shù)介紹  7
2.1  奇異值分解（SVD）的特性   7
2.2  Arnold變換   8
2.2.1  Arnold變換及其周期性  8
2.2.2  Arnold反變換算法   9
2.3 SVD與其他算法的結(jié)合  10
2.3.1  基于DCT和SVD的數(shù)字圖像水印技術(shù)  10
2.3.2  基于DWT和SVD的數(shù)字水印算法   11
2.3.3  基于DFT－SVD域抗幾何攻擊圖像水印算法   12
2.4 數(shù)字圖像處理及Matlab簡介   14
2.4.1  Matlab簡介  14
2.4.2  數(shù)字圖像處理簡介   16
第三章  基于奇異值分解的數(shù)字圖像水印算法   18
3.1  水印嵌入算法  18
3.2  水印提取算法  18
3.3 改進(jìn)的基于SVD的數(shù)字圖像水印算法   20
3.4 攻擊測試與結(jié)果分析  22
3.5  嘗試與離散小波變換結(jié)合  30
3.5.1  算法介紹  30
3.5.2  攻擊測試  30
第四章 總結(jié)  35
4.1 工作總結(jié)  35
4.2 展望  36
致謝   37
參考文獻(xiàn)   38
附錄   40

摘 要

數(shù)字水印是將身份確認(rèn)信息或保密信息鑲嵌于圖像中的一種技術(shù)，可靠的水印可為信息的安全提供可靠的保證。目前許多水印算法是在空間域或變換域插入數(shù)據(jù)的，例如離散余弦變換（DCT）、離散傅立葉變換（DFT）、離散小波變換（DWT）。在本文中，采用了基于奇異值分解（Singular Value Decomposition）的數(shù)字水印算法。圖像奇異值分解（SVD）有以下性質(zhì)：分解后圖像矩陣的奇異值集中反映了圖像的“亮度”（能量）特性，而對應(yīng)的奇異矩陣只反映了圖像的“幾何”特性。因而奇異值的細(xì)微變化不會影響圖像的視覺效果。
在本文算法中，置亂用于數(shù)字圖像隱藏的預(yù)處理和后處理。對數(shù)字水印信號進(jìn)行置亂分散了原始水印信號的相關(guān)性，在遭到剪切攻擊時可以將錯誤碼元盡可能分散，因此有效地提高了數(shù)字水印算法的抗剪切攻擊性能。本算法是采用Arnold變換對水印圖像進(jìn)行置亂的，但利用Arnold變換周期性來恢復(fù)原圖的計算量很大。所以在后處理過程中，采用了一種利用逆變換矩陣來求Arnold反變換的算法。本文算法還對提取出的水印進(jìn)行了量化。首先確定像素值為1的下限和上限，然后對提取的水印圖像進(jìn)行了二值化處理，使最終的水印圖像效果更佳。本論文算法還嘗試將奇異值分解與離散小波變換相結(jié)合，即將水印嵌入到原圖像二維離散變換后所得低頻部分。實驗結(jié)果表明，基于奇異值分解的本算法對常用的圖像處理攻擊具有良好的魯棒性和不可見性。

關(guān)鍵詞：奇異值分解，數(shù)字圖像水印，魯棒性，Arnold變換

ABSTRACT

Digital watermarking is a technique that can inlay identity information and secrecy information into images. Reliable watermarking provides a pledge for information safety. Many current watermarking algorithms insert data in the spatial or transform domains like the discrete cosine, the discrete Fourier, and the discrete wavelet transforms. In this paper, propose a digital watermarking algorithm based on Singular Value Decomposition (SVD). According to some properties, of SVD, each singular value (SV) specifies the luminance of the SVD image layer, whereas the respective pair of singular vectors specifies image geometry. Therefore slight variations of SV cannot effect the visual perception.
In the algorithm of this paper, scrambling technology is used as pre-processing and post-processing of digital image information hiding. The image watermarking is permuted to reduce the relativity of original pixels, so the error bits of the extracted watermarking are dispersed as well. Therefore the resistance to crop attack is improved significantly. In the algorithm, use Arnold transformation to scramble the watermark image, but the image resumption is computational expensive due to the periodicity of Arnold transformation. Therefore, in the post-processing process, used one kind of algorithm how ask the Arnold inverse transformation using the athwart transformation matrix. In the algorithm, but also has carried on the quantification to the recovered watermark image. First determined the lower limit and the upper limit of picture element value is 1, then to the recovered watermark image has carried on two values processing, caused the final watermark image effect to be better. In the algorithm, SVD will try to combine with DWT, and watermark is embedded into the low-frequency part of the original image after two-dimensional discrete wavelet. Experimental result show that the watermarking method based on SVD used to attack the image processing performs well in both imperceptibility and robustness.

KEY WORDS  singular value decomposition, digital image watermarking, robustness, Arnold transformation