預測方法是可逆式資訊隱藏中常用的一個方法,然而不同影像因區塊的特性不一,若使用相同的預測方法進行整張影像預測,可能會因此降低了預測的精準度,進而影響影像品質與藏入量。本論文第一個研究方法改良Feng和Fan學者所提出的延伸Lukac方法,利用區塊的特性進行不同的預測方式,從而提高每個區塊的預測精準度,並且將提出方法應用於16位元醫療影像中;第二個研究方法則改良Lee和Huang學者所提出的插值擴張技術,加入邊緣偵測法及鏡射三角形方位概念減少影像的鋸齒狀效應,以利線條明顯的區塊可以有較好的預測精準度,有效提升影像品質與藏入量。實驗結果證實本論文所提方法在不同的影像上皆有較佳的影像品質與藏入量。
The predicting method is a common used method in reversible data hiding, however different image because the characteristic of area piece is different, if use same predicting method to carry on the whole piece of image to predict, may consequently reduce the precise degree of predict, then influence the image quality and embed capability. Therefore, this thesis proposed two research methods to solve this problem. The first method expands Lukac’s method as originally proposed by Feng and Fan for medical image. This study determines what prediction method should be applied based on standard deviation thresholds to obtain more accurate prediction results. The second method improves Lee and Huang’s approach to reduce the jagged phenomena based on median edge detector and the triangular mirroring concept. In experimental results, the proposed method provides efficiency, high quality, and hidden capacity.