利用以預測為基礎的可逆性映射之新資訊隱藏技術與其在視訊監控與偽裝學上之應用

Abstract

隨著數位攝影機技術的快速發展，有越來越多的監控攝影機架設在日常生活環境中，這些視訊監控系統對於個人隱私權所造成的威脅日益受到關注。本論文利用資訊隱藏技術，對監控影像中的隱私區域做保護與偽裝之研究。首先，本論文提出一個新方法可用以保護視訊監控影片中指定隱私區域的影像，該法利用可逆的映射函式，將隱私區域的影像轉變為與背景圖相似的偽裝影像。接著將此方法延伸到保護視訊監控影片中動態的私密影像部位。此一延伸方法，首先會在每一監視畫面中偵測移動事件並自動決定一隱私保護區域，再根據此區域內影像的色調搭配一無接縫的影像接合技術來調整預設背景圖的顏色，使得偽裝影像與原監視畫面的色調一致。在需要恢復原來的隱私影像部位時，亦可利用此可逆性映射函式的反函式，將偽裝影像無失真地轉回原來的隱私影像。最後，本論文運用前述的可逆性映射函式提出一個新的圖像偽裝術。此法可以將一張全彩的秘密圖像依據指定的目標圖像轉換成偽裝圖像，使得偽裝圖像與目標圖像在視覺上很相近，而且此方法在不使用額外的資訊下，仍然可以無失真地將偽裝圖像恢復為秘密圖像。本論文所提出的方法皆可獲致很好的實驗結果，證明其在實際應用上皆為可行。

With the rapid development of digital camera technologies, more and more surveillance cameras are installed in our daily space for various purposes. There is a growing concern that surveillance systems pose threats to privacy. In this study, we integrate information hiding techniques into surveillance systems to propose three methods for the purposes of privacy protection in video surveillance as well as image steganography. At first, a new method for protecting a selected private region in a surveillance video is proposed, in which reversible prediction-based mappings are used to convert a privacy-sensitive image into a camouflage image which looks similar to a user-selected background image. The method is extended further to conceal private motion activities in surveillance videos, by which motion regions are detected automatically as the privacy-sensitive image parts, and natural and seamless private image hiding are conducted. The reversible mappings guarantee lossless retrieval of the concealed privacy-sensitive image part from the camouflage image, when required. Moreover, a new method using the above-mentioned reversible prediction-based mappings for image steganography is proposed, which is capable of hiding a full-color secret image into any cover image, as well as recovering the embedded secret image losslessly without using overhead information. Good experimental results are also presented to show the feasibility of the proposed methods for real applications.