應用於版權保護之小波樹浮水印技術

Abstract

在過去這幾年間，數位浮水印技術已被廣泛地使用在多媒體資料的版權保護和使用權驗證上。在本論文中，我們提出了一個具強健度但不需要原始影像作為解調工具的數位浮水印方法。首先，在編碼的過程中，我們藉由微小改變LSB plane的值，將浮水印遷入小波樹的尾巴部分。此時，人類視覺模組可以使用來改善因加入浮水印而造成的影像的視覺破壞，並使得浮水印更不易被人眼所察覺。為了確保所加入的浮水印更強健與避免容易被破壞，浮水印被分散在從低頻到高頻的區域。在編碼過程中，藉由隨機選擇小波樹，可以確保浮水印具隱密性而不易被破壞。將有意義的文字或圖片通過單向決定函數（One-way deterministic function），可以驗證使用者的擁有權。最後，實驗的結果證明我們所提出的方法能有效地保護所加入的浮水印，避免被嚴重破壞，並確實保護版權使用權。

In the past few years, digital watermarking has been widely used for copyright protection and proof of ownership in digital multimedia data. In this thesis, we want to generate a robust but original data free watermarking scheme. Wavelet trees are first adopted for watermark embedding, where the watermark information is embedded by applying minor modification to LSB (Least Significant Bit) planes. HVS (Human Visual System) can be introduced to improve the visual quality degradation, and make watermark more transparent. In order to keep watermark robust and transparent, modifications are applied to LSB planes from low to high frequency area. Wavelet trees randomly choosing and one way deterministic function will help to keep watermark secret and not easy to be destroyed by counterfeiters. In the end, the experiment results will show that this watermarking system is effective against attacks.