以小波轉換為基礎的影像浮水印與壓縮; Wavelet-based image watermarking and compression

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Computer Science and Information Engineering > Electronic Thesis & Dissertation > Item 987654321/8399

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/8399

Title:	以小波轉換為基礎的影像浮水印與壓縮;Wavelet-based image watermarking and compression
Authors:	謝明興;Ming-Shing Hsieh
Contributors:	資訊工程研究所
Keywords:	壓縮;數位浮水印;小波;模糊濾波器;Teager;EZW;contextual;compression;watermarking;wavelet;Fuzzy inference filter
Date:	2002-01-18
Issue Date:	2009-09-22 11:25:57 (UTC+8)
Publisher:	國立中央大學圖書館
Abstract:	由於網際網路及資訊科技的快速發展，彈指之間就可以發送電子郵件到世界各地，類似的新觀念逐漸改變了人類以往的行為模式，也同時製造了新的問題，例如盜版的猖獗使得著作權備受挑戰。因此，對著作財產權做適當的保護，以免阻礙了創作者公開其作品的意願，或是對於機密的圖資加以隱藏，不易為他人所發掘，已成為相當重要的課題。浮水印是一種資訊隱藏的應用，就是將機密影像或版權商標等嵌入不同的數位媒體，再透過所擷取的原始浮水印，以進行智慧財產權之證明。在網路傳送之中難免會對影像作壓縮等處理，這些處理的過程會對嵌入影像的原始浮水印造成的破壞，例如慣常使用的壓縮、均化、強化等均會使得浮水印的辨識度降低，因此使用於數位影像資訊的浮水印需符合不易察覺(Invisible)、明確性(Unambiguous)、強韌性(Robustness)、抵抗破壞(Tamper-resistance)等特性。本論文提出以多解析度領域為基礎的小波轉換技術以嵌入浮水印，我們嵌入的浮水印均為有意義的影像資訊，例如二值影像、灰階影像、及彩色影像，為了加強嵌入後滿足對原始影像的不可察覺性，以及強化所擷取的浮水印，論文中我們提出六種方法，透過計算選定波段係數能量值的方法，以選擇嵌入的浮水印係數位置，第一種方法為QSWT，所選取的係數必須同時滿足該係數以及其小波子樹都大於特定值；此種選取係數的方法也可以提升影像壓縮的倍數與品質。第二種方法是考慮周邊點對目標點的影響程度。其能量值是計算本身及周邊點大於均值或均值加部分標準差的個數權重值。第三種方法是另外考慮其子樹的均值及標準差。第四種方法我們透過計算亂度的Teager能量運算子計算其周邊點與目標點的亂度值，找出較大的亂度值所對應的係數，以取代其最小或次小位元。第五種方法是以第三種方法為基礎，加上模糊歸屬函數的判斷計算其能量值。第六種方法我們以log-sum去計算所有係數的亂度值，找出較大亂度值所對應的係數，接著對彩色浮水印做三階轉換分頻，以自動調整嵌入強度的方式嵌入浮水印。在實驗中我們以不同類型的影像作為測試的原始影像，也以不同類型的影像作為測試的浮水印，對原始影像的不可察覺性以及擷取的浮水印抗影像處理及抗壓縮的強度上均有優於直接以排序找較大係數嵌入的作法。此外我們也提出以模糊推論濾波器及調適性量化為基礎的影像壓縮方法。我們所使用的模糊推論濾波器是用來判斷小波零樹，並以調適性量化編碼用來增進壓縮品質。從實驗中證明，我們均獲致比EZW及JPEG更好的壓縮品質與壓縮倍數。 Copyright protection and security of image contents are nowadays all-increasing demands especially with the drastic expansion of Internet and web-based services. To address the demands, watermarking methods have been quickly developed. Watermarking methods are classed into two main classes: spatial- and spectral-based approaches. All methods aim at satisfying two basic requirements: watermark perceptual invisibility and watermark robustness against to attacks. Six wavelet-based watermarking approaches are presented in this dissertation. The proposed approaches exploit the spatial localization and frequency spreading of wavelet transform to embed watermarks in the coefficients with larger local energy to achieve high imperceptibility and robustness of watermarks. The first proposed approach embeds binary or gray-level watermarks in images by modifying the wavelet coefficients of a host image. A multi-energy watermarking scheme based on the qualified significant wavelet tree (QSWT) is used to select embedding coefficients. QSWT can also be adopted for image compression and gets better results than EZW and JPEG get. Two watermark embedding strategies which consider the local characteristics to choose the larger-energy DWT coefficients for embedding bi-level or gray-level watermarks are then proposed. In the second approach, the embedding coefficients are selected according to the weighted contextual energy of the coefficients in a DWT subband. In the third approach, the embedding coefficients are selected according to the weighted contextual energy defined on the considered subband and all its descendant subbands. The proposed approaches have no need of the original host image to extract watermarks. The fourth approach is based on the Teager energy operator considering the local characteristics to choose the larger-energy DWT coefficients in the low-frequency subband to embed bi-level watermarks. The watermark is pre-encrypted by a random binary sequence for security before embedding; the embedding coefficients and the watermark need not be sorted. The embedding process is done by replacing the LSBs or the second LSBs of the wavelet coefficients with the encrypted watermark to raise the security. The fifth watermarking technique is based on the context measurement and fuzzy filter. The fuzzy filter is employed to conclude the context of each coefficient. The embedding coefficients have their own embedding degrees which are automatically calculated in accordance with the energy contributed from the context. The original host image is not required to extract watermarks. The sixth watermarking framework embeds color watermarks in color images, which can resist image-processing attacks, such as JPEG compression, JPEG 2000 compression, and composite image processings. The embedding locations were selected based on a contextual energy measurement. An adaptive casting strategy was proposed to embed watermark coefficients for completely controlling the imperceptibility of watermarked images and the robustness of watermarks. The proposed approach has no need of the original host image to extract watermarks. Other than watermarking approaches, we also proposed a subband image coder with fuzzy inference filter and adaptive quantization. By modeling the DWT subbands, the proposed fuzzy inference filter instead of Shapiro’s uniform thresholding was used to calculate coefficient context to determine zerotree roots. The adaptive quantization was used to improve the fuzzy classification performance. The experimental results show that the proposed approach is superior to EZW and JPEG; especially, in the cases of high-ratio compression.
Appears in Collections:	[Graduate Institute of Computer Science and Information Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Size	Format

社群 sharing

Loading...