於本論文中,我們重新探討基於預測差異擴張法技術之可逆的資料隱藏技術。近期,Li等人提出一個新的預測差異擴張法技術,使用它去建構一個高品質可逆的資料隱藏方法。然而,他們提出方法的隱藏容量都不是相當高。為了提高容量,可以直接使用多層嵌入的方式增加容量,但卻會造成大量圖形品質降級。 在我們的論文當中,我們先提出一般化之預測差異擴張方法,我們稱為S-max、S-min演算法。用這兩個演算法,我們可以簡單地刻畫Li等人提出的預測差異擴張方法。為了防止Li等人提出的方法在多層運行中降級,我們也提出了Max-Min-Min-Max的運作概念,運用在資料隱藏上的技術。 最後,實驗結果將展現出我們提出的方法比起Li等人提出的兩層嵌入方法可獲得更多的藏量與更高的圖形品質。
We revisit reversible data hiding schemes based on prediction-error-expansion techniques. Recently, in Signal Processing, Li, Li, Li, and Yang proposed a new prediction-error-expansion technique and used it to construct a high-fidelity reversible data embedding scheme. However, the embedding capacity of their proposed scheme is not high. Directly applying multi-pass embedding increases the embedding capacity but results in a large of image degradation. In this paper, we generalize the prediction-error-expansion technique proposed by Li, Li, Li, and Yang. Then, we use the generalized schemes to characterize their proposed data embedding scheme. In order to prevent the image degradation due to applying two-pass embedding, we propose a novel data embedding scheme which is a combination of our proposed generalized data embedding schemes. Finally, experimental results show our proposed scheme obtains higher embedding capacity and image quality than the two-pass embedding scheme proposed by Li, Li, Li, and Yang.