以隱藏式商標保護影像所有權之研究

Abstract

隨著數位多媒體的快速發展，使得數位資料得以快速簡便地複製與傳遞，因此電子文件或圖片的著作權的保護也越來越重要。數位浮水印技術可以利用嵌入看不見的訊號於多媒體資料中來驗證資料擁有人的身份，並且防止未經授權的複製。本論文提出三種在偵測浮水印時不須原始影像的數位浮水印技術。第一種方法屬於空間域上的數位浮水印技術，我們利用直接修改影像中某些圖素的灰階值來嵌入視覺上察覺不出的圖案。第二種方法屬於頻率域上的數位浮水印技術，我們藉由修改在中頻區域的係數來嵌入我們的圖案。第三種方法則是我們結合前面兩個方法的優點，提出的一個混合的數位浮水印技術，以就是浮水印同時出現在影像的空間域和頻率域中。根據實驗結果顯示，我們所提出的方法對於各種對影像的更改，例如：失真的資料壓縮、影像的割除、加雜訊、模糊化，以及對比度增強等，都有不錯的效果。

With the fast growth of digital media, the digital data transmission and reproduction become quick and easy. Therefore, the copyright protection of digital media becomes increasingly important. Digital watermarking technique has been proposed as a method to embed an invisible signal into multimedia data in order to attest the owner identification and discourage the unauthorized copying. In this thesis, we propose three digital watermarking schemes which do not require original image for watermark detection. In the first approach, we embed digital watermarks by directly modifying the intensity of certain pixels with visually imperceptible patterns hidden in the images. In the second approach, we change our approach from the spatial domain to the frequency domain in order to design a watermarking scheme which can survive after lossy data compression. We embed the digital watermark by modifying the middle-frequency coefficients of the image. In the third approach, we combine the advantages of the first method with the second to propose a more robust watermarking technique. The watermark is embedded into each of the spatial and frequency domains of image. Experiment results show that this method could be more robust to resist several kinds of image processing attack including lossy data compression, image cropping, noise addition, image smoothing, and contrast enhancement.