利用小波轉換進行浮水印加密之研究

在本研究中，我們提出一套數位水印加密演算法，此演算法在一灰階影像內藏入影像大小四分之一的半色調像，最終並能成功將其解密。首先我們利用Haar小波轉將此一灰階影像分解成三層多解析度影像，再將半色調影像分解成二層多解析度影像，之後利用pseudo-random number將多解析度半色調數位浮水印重新排列，再依其不同解析度來做影像藏入，在解密時；我們必須要有已加密的影像與加密時所採用的特定seed值，方可成功解密。在加密演算法強韌性的評估上，我們亦嚐試了許多常見的數位影像處理技術來予以破壞，但最終我們皆能加以成功解密。驗結果顯示本研究所發展的演算法具有相當好的強韌性。

In this research, a watermarking technique based on the multiresolution structures of wavelet decomposition is proposed. The main purpose is to hide a binary watermark into a multiresolution representation by Haar wavelet transform, and a binary watermark, whose size is a quarter of that of the host image, is decomposed into another multiresolution representation. We permute the decomposed watermark by using pseudo-random number. Each layer of the resulted watermark is embedded into the corresponding layer of the decomposed host image. In decoding, the seeds and the encoded host image are all necessary to successfully retrieve the binary watermark. We use several common methods of digital image processing to attack the encoded image and then estimate the embedded watermark. Experimental results demonstrate that our proposed watermarking technique is of pretty good robustness and validity.