外文翻譯及文獻(xiàn)--基于子帶離散余弦變換dct應(yīng)用于圖像水印的技術(shù)

1、<p><b>　　中文2350字</b></p><p><b>　　畢業(yè)論文（設(shè)計(jì)）</b></p><p><b>　　外文翻譯</b></p><p>　　題　　目：　　　基于DCT變換的水印算法實(shí)現(xiàn)　　　</p><p>　　?！　I(yè)：　　　　　　　　　&l

2、t;/p><p>　　班　　級(jí)：　　　　　　　　　　　　</p><p>　　學(xué)　　號(hào)：　　　　　　　　　　　</p><p>　　姓　　名：　　　　　　　　　　　　　　　</p><p>　　指導(dǎo)教師：　　　　　　　　　　　　　</p><p>　　基于子帶離散余弦變換DCT應(yīng)用于圖像水印的技術(shù)</p>&

3、lt;p>　　基于子帶離散余弦變換DCT應(yīng)用于圖像水印的技術(shù)已經(jīng)被提出并應(yīng)用。水印是波在所有選定的含有若干系數(shù)的四個(gè)頻帶段的1級(jí)分解。應(yīng)用大量的系數(shù)使每個(gè)波段給出了不同的檢測(cè)輸出結(jié)果。其結(jié)果是采取平均檢測(cè)結(jié)果的所有頻段的值。結(jié)果表明，最終的結(jié)果是優(yōu)于所檢測(cè)輸出的每個(gè)波段所得的結(jié)果的，從而實(shí)現(xiàn)了非常強(qiáng)大的水印方案。</p><p><b>　　1、導(dǎo)言</b></p>&l

4、t;p>　　數(shù)字媒體技術(shù)在當(dāng)今社會(huì)已被大范圍的使用，從而促使其創(chuàng)立知識(shí)產(chǎn)權(quán)來(lái)保護(hù)。就其性質(zhì)而言，數(shù)字媒體是能夠100％被完整復(fù)制的，因此，必須采取有效的標(biāo)識(shí)系統(tǒng)（是顯而易見(jiàn)的）。這就是水印的由來(lái)。水印技術(shù)是指將無(wú)法被看見(jiàn)的數(shù)據(jù)埋入圖像中，從而確定合法的創(chuàng)建者/擁有者。水印應(yīng)當(dāng)具有健全的可以適用于（抵擋）各種各樣的圖像攻擊的技術(shù)。任何嘗試從原始圖像刪除所有權(quán)信息的方法（被稱為）攻擊。一些常見(jiàn)的攻擊包括過(guò)濾，壓縮，直方圖修改，剪裁，旋

5、轉(zhuǎn)和縮小。（水?。┲饕袃蓚€(gè)嵌入方向，即空間域和變換域。變換域的技術(shù)對(duì)普通的圖像攻擊技術(shù)更敏感，如過(guò)濾或JPEG壓縮。</p><p>　　變換域技術(shù)在圖像水印中是最受歡迎的。在這種情況下，圖像技術(shù)正在通過(guò)某些常見(jiàn)的，頻繁發(fā)生事情改變著，并且使得水印轉(zhuǎn)換系數(shù)被高度完美的應(yīng)用于圖像上。這種轉(zhuǎn)換技術(shù)通常使用DCT（離散余弦變換）,DFT（二維傅里葉變換）和DWT（離散沃爾什變換）。</p><p&

6、gt;　　發(fā)生在現(xiàn)況下的一個(gè)問(wèn)題是各種數(shù)量和位置的改變將使其在圖像中頻繁的變化著。許多有效方法已經(jīng)被提出，其中大部分是源于科克斯（Cox’s）的體系。皮瓦等人擴(kuò)展了這一方法，從而提出了一種隱藏檢測(cè)系統(tǒng)（blind detection system）。在這些全部情況中都將圖像處理作為一個(gè)整體，但一些系數(shù)變化不超過(guò)16000，通常的圖像尺寸是512x512。由于大多數(shù)的過(guò)程都是在數(shù)字統(tǒng)計(jì)的背景下（執(zhí)行的），因此我們寧愿使用越多系數(shù)越好。這就

7、是為什么我們建議使用子帶離散余弦變換DCT原因。</p><p>　　第2節(jié)，我們將圍繞目前的子帶DCT分解層和模型的參數(shù)進(jìn)行討論。第3節(jié)，我們將所閱讀到的方案進(jìn)行測(cè)試，并且解釋每一個(gè)波段的處理后的情況下，最后檢測(cè)未經(jīng)過(guò)處理的五個(gè)通常攻擊方法。最后，我們將在最終章節(jié)結(jié)束這一討論。</p><p>　　2、子帶DCT和水印的模型</p><p>　　鄭和米特拉（Jun

8、g and Mitra）已經(jīng)于1996年介紹了子帶DCT。這是一種涉及小波變換和離散余弦變換（DCT）的方法。將原始圖像二次抽樣后經(jīng)過(guò)高通濾波過(guò)濾器和低通濾波過(guò)濾器（處理）。結(jié)合這兩個(gè)過(guò)濾器的各個(gè)方向（橫向和縱向）的過(guò)濾使四個(gè)子帶為每個(gè)層進(jìn)行分解。這種過(guò)程相當(dāng)于通過(guò)低通濾波器向各個(gè)方向透進(jìn)的頻帶能夠進(jìn)一步的被二次抽樣和過(guò)濾，使其能在另一層被分解。最后，使每一個(gè)頻帶都通過(guò)使用DCT系數(shù)來(lái)進(jìn)行轉(zhuǎn)化。</p><p>

9、　　在我們的實(shí)驗(yàn)中，我們不得不選取并使用一定數(shù)量的的分解方法和小波技術(shù)。我們嘗試用大量的分解層，即1，2和3層來(lái)進(jìn)行試驗(yàn)。實(shí)驗(yàn)結(jié)果表明，(分解系數(shù))并沒(méi)有在某一層的探測(cè)結(jié)果上有重大的提高，與此同時(shí)，圖象退化的現(xiàn)象卻更容易被發(fā)現(xiàn)。因此，我們把已經(jīng)被分解的原始圖像的四個(gè)頻段放入一層內(nèi)。這種嫻熟的使用最簡(jiǎn)單的小波技術(shù)的方法，是哈爾（Haar）提出的。下一步是將每一頻段進(jìn)行DCT變換。為了解釋我們的水印技術(shù)，我們使用下列公式：</p>

10、;<p>　　其中ti是正在轉(zhuǎn)化的系數(shù)，ti’是水印系數(shù)，xi是一個(gè)被用于水印中的隨機(jī)序列的高斯分布。參數(shù)a是與模型的濃度有關(guān)。我們對(duì)其使用兩種不同的準(zhǔn)則，一種是LL-頻段，剩下的將使用另一種頻段,那就是當(dāng)a=0.1的時(shí)候使用LL-頻段，a=0.2時(shí)則用另一種。這樣做的理由是，低頻段更容易變化，也就是說(shuō)，一些細(xì)微的變化更加的明顯。i參數(shù)的范圍是從1到20000，使其在一個(gè)令人滿意的80000系數(shù)間的變化。在每個(gè)頻段中，我們

11、都將從以5000為系數(shù)的鋸齒形依次掃描。</p><p>　　基于塊分類(lèi)和DCT域的圖像水印算法</p><p>　　摘要：本文提出了一種基于離散余弦變換（DCT）域圖像水印算法。</p><p>　　圖像水印算法有兩個(gè)階段：特征嵌入和特征檢測(cè)。第一階段，它將一個(gè)標(biāo)識(shí)符號(hào)嵌入進(jìn)圖像。第二階段是被公認(rèn)（已知）的。該算法有兩個(gè)處理步驟。第一步無(wú)疑是選擇像素區(qū)塊并使用參數(shù)

12、進(jìn)行設(shè)置，而第二個(gè)步驟是將DCT系數(shù)強(qiáng)制的嵌入在選定的區(qū)塊內(nèi)。兩種不同的參數(shù)規(guī)則表明修改DCT參數(shù)系數(shù)出現(xiàn)頻率的重要性。第一種方法是將DCT規(guī)則嵌入到選定的線性約束內(nèi)，而第二種方法則是按照所給予的特定參數(shù)進(jìn)行循環(huán)檢測(cè)。上述所提到的水印算法是不能在JGEG壓縮和過(guò)濾條件下使用的。</p><p><b>　　1、導(dǎo)言</b></p><p>　　數(shù)字水印是當(dāng)今廣播電視和

13、密碼技術(shù)的探討的大體背景下產(chǎn)生的。為了避免（他人）未經(jīng)授權(quán)就發(fā)布圖片或其他多媒體資源，已經(jīng)提出了大量的解決方法。其中多數(shù)是提議做一些難以被發(fā)現(xiàn)的圖片修改以供以后使用。這種圖片修改技術(shù)被稱為水印。水印是將圖片做一些不明顯的修改（以確定版權(quán)所有），從而能夠強(qiáng)力的抵制可能出現(xiàn)的各種圖像處理技術(shù)。</p><p>　　水印算法已經(jīng)被大量的頒布過(guò)。他們不是隨機(jī)性的就是確定性的。這些算法，包含了圖像強(qiáng)度域和變換域。在中頻范圍

14、內(nèi)DCT的變換系數(shù)受嵌入的8*8像素塊所約束。將授權(quán)信息嵌入DCT系數(shù)后應(yīng)用所得的DCT系數(shù)來(lái)處理整個(gè)圖像。</p><p>　　圖像水印算法有兩個(gè)階段：特征嵌入和特征檢測(cè)。通過(guò)特征嵌入來(lái)編寫(xiě)絕對(duì)代碼分配給所有者后讓其嵌入圖像。在檢測(cè)階段用算法來(lái)確定所規(guī)定的代碼。信號(hào)檢測(cè)理論是一種對(duì)許多領(lǐng)域都有效的應(yīng)用技術(shù)。水印圖像能夠用許多不同的處理方法來(lái)轉(zhuǎn)變圖像和處理運(yùn)算法則來(lái)防止其被摧毀，這就是數(shù)字水印技術(shù)。圖像壓縮是每個(gè)

15、圖像都有可能經(jīng)歷的圖像變化過(guò)程。標(biāo)準(zhǔn)的靜態(tài)圖像壓縮算法是JPEG格式。JPEG格式是基于盡可能減少資源在離散余弦變換（DCT）域上的消耗（而產(chǎn)生的）。受損壓縮，即使圖像的信息遭受損失的壓縮方法是在高頻域上發(fā)生的。</p><p>　　在擬定的水印算法中，圖像被分割成類(lèi)似JPEG算法的8*8像素塊。該水印算法包括兩個(gè)步驟。第一步是依據(jù)高斯網(wǎng)來(lái)選擇某些特定的塊。在選定的塊中，我們通過(guò)修改DCT參數(shù)來(lái)使其強(qiáng)制完成某一給

16、予的約束。該參數(shù)是把高斯函數(shù)加在系統(tǒng)規(guī)定的DCT系數(shù)上使其組成水印代碼。在檢測(cè)階段我們首先檢測(cè)DCT參數(shù)，然后檢測(cè)各自的區(qū)塊的位置來(lái)確定是否被篡改。</p><p>　　A SUBBAND DCT APPROACH TO IMAGE WATERMARKING</p><p>　　A subband-DCT approach for image watermarking is propose

17、d in this communication. The watermark is casted in a selected number of coefficients of all four bands of a one-level decomposition. A great number of coefficients is being used. Each band gives a different detection ou

18、tput. The result is taken as the average detection result of all bands. It is shown that the final result is better than the detection output of each individual band, thus leading to a very robust watermarking scheme.<

19、;/p><p>　　1. INTRODUCTION</p><p>　　The great spread of digital media in nowadays, has urged for the protection of the intellectual property rights of the creators. By their nature, digital media ar

20、e 100% reliably copied, so the need for an effective marking system is obvious. This is where watermarking comes in. Watermarking stands for the embedding of perceptually invisible information into image data that identi

21、fy the rightful creator/owner. Watermarks should be robust to various image attacks. Every attempt to remove the owner</p><p>　　The frequency-domain approaches are the most popular for image watermarking. In

22、 these schemes, the image is being transformed via some common frequency transform and watermarking is achieved by altering the transform coefficients of the image. The transforms that are usually used are the DCT, DFT a

23、nd the DWT. A question that occurs in such approaches is the number and the position of the altered coefficients in the frequency representation of the image. Many different ideas have been presented, </p><p&g

24、t;　　In section 2, we present the subband DCT, the decomposition levels and discuss the casting scheme and parameters. In section 3, we test the reading scheme and examine each band’s individual contribution in the case o

25、f no processing and also the final detection results for five common attacks. We end with the final conclusions in section </p><p>　　2. SUBBAND DCT AND WATERMARK CASTING</p><p>　　Jung and Mitra

26、have introduced the subband DCT in 1996. It is a method that involves both wavelets and the Discrete Cosine Transform (DCT). The original image is sub sampled and filtered with the use of a high pass and a low pass filte

27、r. The combination of the two filters for each direction (horizontal and vertical) of filtering gives four subbands for each level of decomposition. The band that corresponds to low pass filtering in both directions (LL

28、band) can be further subsampled and filtered </p><p>　　For our experiments, we had to select the number of decomposition levels and the wavelet to be used. We tried different numbers of decomposition levels,

29、 namely 1, 2 and 3. Experimental results have shown that there wasn’t any significant improvement in the detection results for more levels than one, while at the same time, the image degradation was more easily observed.

30、 So we decomposed the original image into one level with four bands. This was accomplished using the simplest wavelet, that is H</p><p>　　where ti are the transformed coefficients, ti’ are the watermarked co

31、efficients and xi is a random sequence of Gaussian distribution, used as the watermark. The a- parameter has to do with the strength of the casting. We use two different values for it, one for the LL-band and a different

32、 one for all other bands, that is a=0.1 for the LL- band and 0.2 for the others. The reason for this is that the low frequency band is more vulnerable to changes, meaning that slight changes are easily noticeabl</p>

33、;<p>　　Image watermarking using block site selection and DCT domain constraints</p><p>　　Abstract: In this paper we propose an image watermarking algorithm based on constraints in the Discrete Cosine

34、Transform (DCT) domain.</p><p>　　An image watermarking algorithm has two stages: signature casting (embedding) and signature detection. In the first stage it embeds an identifying label in the image. This is

35、 recognized in the second stage. The proposed algorithm has two processing steps. In the first step certain pixel blocks are selected using a set of parameters while in the second step a DCT coefficient constraint is emb

36、edded in the selected blocks. Two different constraint rules are suggested for the parametric modification</p><p>　　1. Introduction</p><p>　　Digital watermarks in the general context of TV broad

37、casting and cryptology were discussed in. To avoid the unauthorized distribution of images or other multimedia property, various solutions have been proposed. Most of them make unobservable modifications to images, that

38、can be detected afterwards . Such image changes are called watermarks. The watermark should not alter visibly the image and it should be robust to alterations which may be caused by various image processing techniques.&l

39、t;/p><p>　　Algorithms proposed for watermarking have been reported in various papers. They are either stochastic or deterministic. These algorithms are either in image intensity domain or in frequency domain. I

40、n the middle range DCT frequency coefficients from the 8*8 pixel blocks are used for embedding a constraint. In the signature is embedded in the DCT coefficients obtained after applying the DCT transform in the entire im

41、age.</p><p>　　An watermarking algorithm has two stages: watermarking casting and detection. By means of watermark casting a specific code assigned to the owner is embedded in the image. In the detection stag

42、e the algorithm identifies the given code. Signal detection theory is a well-established field with many applications. A watermarked image can be processed by means of various image transformations and processing algorit

43、hms which may be able to destroy, intentionally or not, the digital watermark. Image co</p><p>　　In the proposed watermark casting algorithm from, the image is partitioned in 8*8 pixel blocks similar to the

44、JPEG algorithm. The watermarking algorithm consists of two steps. The first step selects certain blocks according to a Gaussian network. In the selected blocks we modify DCT coefficients such that they fulfill a given co

45、nstraint. The parameters of the Gaussian functions and of the imposed constraints on the DCT coefficients make up the watermark code. In the detection stage we first chec</p><p>　　In Section 2 we propose a t

46、echnique for choosing block sites. The DCT constraint embedding step is explained in Section 3. The detection stage is presented in Section 4. This analysis is necessary in order to determine the suitable watermark param

47、eters such that each watermark is distinctly identified from all the others. The simulation results for applying the proposed algorithms in gray level and color images are provided in Section 6. In Section 7 the conclusi