可逆式資訊隱藏可將偽裝影像無失真的還原,因此可逆式資訊隱藏法可用於軍事、醫療、犯罪偵查等高敏感度的影像。可逆式資訊隱藏雖具有較高的影像品質,然而機密訊息的藏入量較低。典型的可逆式資訊隱藏技術可分為差異擴張或是直方圖位移修正法,但在2013年Li等學者提出像素排序法(PVO)其方法簡易且平均影像品質具有52dB之高,與差異擴張或是直方圖位移修正法相較下,像素排序法擁有較高的影像品質,但是藏入量卻較為低。因此為了改良PVO方法藏量不彰,本研究提出不受區塊束縛以及參考像素值束縛的廣義化像素排序法(GePVO)。GePVO方法利用參考像素值往最大像素值或是最小像素值方向逐一進行像素值預測,以促使預測誤差值增加而能有更多的藏入量。廣義化的像素排序法可依參考像素位置不同以及不同的區塊大小而讓原始PVO方法藏入量提升。本實驗中以 12張影像進行實驗,當參考像素位置在區塊的邊界,藏量平均可達70,346位元,其中以Airplane影像具有最高的藏入,與PVO方法藏入量的約3倍,且平均影像品質為40.24dB。然而,當參考像素值為中間像素值時,其藏入量比原始PVO高出1.8倍且平均影像品質可高達48.09dB。可見,本研究提出廣義化的PVO方法(GePVO),具有高負載容量及高影像品質的特性使得GePVO可順應各種應用的可調性。
Stego image can be recover to original state by reversible data hiding technology which applicable to military, medical, high-crime investigation and sensitivity images. Reversible data hiding with higher image quality but embedded capacity is lower. The classic of reversible data hiding are two main categories. Histogram shifting and modification and difference expansion. In 2013 Li et al. proposed pixel value ordering method which is easy to use and the stego image quality could be 52dB high. Compare with histogram shifting and modification and difference expansion, pixel value ordering method with the highest stego image quality. But most of reversible data hiding method with lower embedding capacity. This research propos a general pixel value ordering (GePVO) method which is unshackle of block size and reference pixel to preduce. GePVO uses reference pixel to the maximum pixel and minimum pixel to predict each pixel. With more prediction error could increase more embedding possibility. General pixel value ordering method in difference reference pixel and difference block size could make difference result of embedding capacity which is higher than original pixel value ordering method. In this research we use 12 image for experiment. The embedding capacity 70,346bit on average when reference pixel in the bound of block. The embedding capacity is higher than PVO method 3 times on average and the image quality maintain in 40.24dB on average. However the embedding capacity is higher than PVO method 1.8 times on average when the reference pixel in the middle of the block. The image quality maintain 48.09dB on average. In this research propos a general pixel value ordering method (GePVO) which has high embedding capacity and high image quality to improve GePVO with more adaptive and adjustable.