灰階影像藏匿彩色浮水印採用（2,2）門檻秘密分享策略

Color Watermarking for a Grayscale Image Based on (2,2) Threshold Secret-Sharing Schemes

隨著資訊科技的日新月異，數位資料的編輯、修改以及複製的方便性，使得如何保障產權就變成十分重要的議題。Visual secret sharing strategy是基於數位影像產權所產生的一種隱藏式浮水印技術。在過去，基於視覺密碼模型之浮水印技術，對於嵌入的浮水印僅黑白或者是簡單的幾個顏色文字。因此本文研究採用（2,2）門檻秘密分享策略，實現灰階圖像嵌入彩色的浮水印並結合Toral automorphism來達到對於影像攻擊如剪裁、模糊、雜訊、銳化、JPEG壓縮的強韌性，在實際實驗我們以剪裁10%、20%、30%、模糊化、雜訊、銳化、JPEG QF=90、50、10作為影像攻擊手段。實驗結果表示，裁20%與30%攻擊下之還原浮水印NC值分別為0.857與0.808，其餘攻擊的NC值皆在0.934以上。最後，增加水平/垂直的扭曲攻擊與旋轉攻擊實驗，本研究所提出的方法，能夠有效抵禦水平/垂直扭曲攻擊以及15度以下的旋轉攻擊。

Computer and communication technology has increased the transmission and exchange of digital data and created an environment in which digital information is easy to spread, copy, and modify, necessitating effective copyright protection techniques. Previous watermark studies have used only binary, grayscale, or a four-color images for visual cryptography. In this study, the RGB colorwatermark was embedded into a grayscale image based on (2, 2) threshold secret-sharing schemes. We propose two strategies: combining binarzation and toral automorphism, and combining “neighboring relations” and toral automorphism. In the experiment, the attack resilience was evaluated it contained 10% clipping, 20% clipping, 30% clipping, blurring, noise, sharpening, scaling, JPEG QF = 90, JPEG QF = 50, and JPEG QF = 10. The neighbor relations approach reduces the destruction of the image that is embedded information. The watermark NCs under 10% clipping and 20% clipping attacks were 0.857 and 0.808, and the remaining NCs exceeded 0.934. Finally, we added horizontal/vertical distortions and rotation attack to the experiment. The results show that the proposed method effectively resists horizontal/vertical distortion and rotation attacks below 15°.