植基於像素排序法之可回復資訊隱藏技術

本研究是基於Weng等學者於2017年提出的最佳化（Optimal）基於像素之像素排序法（Pixel-based Pixel Value Ordering, PPVO）及Wu等學者於2020年提出的改良式方法（Improved PPVO），來建構出一套可回復式資訊隱藏方法，在提取秘密訊息的同時，可以還原出原始影像。本方法可在人眼難以察覺之細微變化下，於灰階影像中藏入秘密訊息，並在低藏密量的前提下，有效提昇藏密品質，使之更為隱密。本方法將載體影像以棋盤式切割成兩部分，再各自獨立進行PPVO藏密，並依照PPVO編碼像素的區塊複雜度（光滑程度），來適應性地調整參照區塊大小。於測試驗證上，以BOSSBase自然影像資料庫中之載體影像為例，在各種藏密量的設定下，所獲得之影像品質普遍優於Weng等學者提出的最佳化PPVO方法和Wu等學者提出的改良式PPVO方法。本方法並可抵抗正規／特異（Regular-Singular, RS）分析和像素差直方圖分析，並在廣泛性測試有良好的表現。

關鍵字

資訊隱藏；可回復資訊隱藏；適應性像素修改； PPVO

並列摘要

In this study, a reversible image steganographic scheme that is able to restore original cover images after retrieving secret information is developed based on the optimal pixel-based pixel value ordering (PPVO) by Weng et al. (2017) and the improved PPVO by Wu et al. (2020). By this method, secret information is embedded in grayscale images and only causes minor changes imperceptible to human vision. The quality of stego images is enhanced in the case that a certain low amount of information is embedded. The cover image is split into two parts interleavedly in a checkerboard-like manner, and each part performs the embedding process independently. With different pixel complexity, the size of context pixels is adaptively modified. During experimental verification, the images in BOSSBase dataset are adopted as the cover images. The quality of stego images is generally better by using this method, compared with the one by Weng et al. (2017) and Wu et al. (2020). This method can also prevent the stego images from being detected by the regular-singular (RS) analysis and the pixel difference histogram (PDH) analysis, and performs well during generalized benchmarking.

並列關鍵字

steganography ； reversible image steganography ； adaptive pixel-modification ； PPVO

主題瀏覽