Abstract: | 應用於加密影像的可回?式資訊隱藏技術是一個熱門的研究議題,此項技術?但可以利用加密演算法強化原始影像的隱密性,又可透過可回?式資訊隱藏機制將額外訊息嵌入至加密影像中,提供訊息傳遞、資?管?或存取控制等時機使用。然而,隨著三維模型及高動態範圍影像相關應用的迅速發展,適用此兩種加密媒體的可回?式資訊隱藏技術仍是一個亟待研究的議題。本文探討應用最高有效位元預測與霍夫曼編碼於加密三維模型與加密高動態範圍影像可回復式資訊隱藏技術的可行性。在應用於加密三維模型的技術中,我們先將輸入模型的頂點座標值有效地轉換成介於0與1之間的小數,使得後續最高有效位元預測技術能順利執行。此外,也透過隨機鄰居挑選技術增加可嵌入頂點的數量,以大幅提升訊息總嵌入量。最後,我們妥善規劃輔助訊息的處理方式,完善現有技術的運作流程。在應用於加密高動態範圍影像的技術中,我們利用Radiance RGBE影像格式中相鄰像素E頻道數值的高相似性,使用最高有效位元預測技術大幅提升訊息嵌入量,並利用自訂的影像加密技術保有此影像格式的特性後,再透過同質索引值修改進行訊息的二次嵌入。實驗結果顯示,本文所提出的兩項技術能擁有高訊息嵌入量、高嵌入率、可回復性與可分離性等特性,有效拓展可回復資訊隱藏技術於?同加密媒體上之應用。 Reversible data hiding in encrypted images (RDHEI) is a popular research topic. This technique can not only enhance the privacy of the original image using encryption algorithms, but also embed additional data into the encrypted images through reversible data hiding for the purpose of covert communication, data management, and access control. With the rapid development of other applications on 3D models and high-dynamic-range (HDR) images, reversible data hiding in these two encrypted media is still a problem to be solved.This study proposes two reversible data hiding schemes for encrypted 3D models and encrypted HDR images separately based on most-significant-bit (MSB) prediction and Huffman coding. For the algorithm in encrypted 3D models, we first convert the vertex coordinates of the input model into decimal values between 0 and 1 to effectively perform MSB prediction. Then, random sampling is integrated to select partial neighboring vertices of the processing one to increase the total embedding capacity. The number of the embeddable vertices can be also significantly raised. Thereafter, we properly handle the side information to complete each process of the proposed algorithm. For the algorithm in encrypted HDR images, we make good use of the high similarity of E values between neighboring pixels in the HDR image with the Radiance RGBE image format. Thus, through MSB prediction and Huffman coding, higher embedding capacity can be achieved. We also propose a novel image encryption to keep the characteristics of the HDR image with the Radiance RGBE image format. Thus, a lossless data hiding algorithm called the homogeneity index modification can be integrated to improve the embedding capacity. The experimental results demonstrate the feasibility of the proposed two algorithms with the characteristics of higher embedding capacity, higher embedding ratio, reversibility, and separability. Keywords: Reversible data hiding, Encrypted media, MSB prediction, Huffman coding, 3D models, High-dynamic-range images |