可供普及溝通之低扭曲「富含訊息視訊」

Abstract

在人們的生活周遭，存於環境中許多種類的物體都可用來容納資訊，以達到普及溝通(pervasive communication)的目的。廣義上來說，這些物體皆可被視為多媒體，會以不同形式出現在生活周遭，如廣告看板、海報、在螢幕上播放的視訊等等。訊息可藉由資訊隱藏的技術嵌入其中，而這些隱藏的訊息更可藉由手機、平版這些智慧型行動裝置抽取出來，此種隱含訊息的多媒體即稱為「富含訊息多媒體(message-rich multimedia)」。 本研究提出一新技術來製作富含訊息的視訊，與以往的訊息傳遞方式不同，人們可藉由「放映與再取像(display-and-imaging) 」的方式來抽取藏入的訊息，而不必透過直接傳輸數位檔案的方式交換訊息。此外，這個新製作出的富含訊息視訊，其畫質相較於先前提出的方法 [26] 有較低的扭曲。為了達到這個目的，本論文設計了訊息嵌入的程序以及一個對應的訊息抽取程序。 在訊息嵌入程序中，本研究採用了嵌入「對角位峰值訊號(diagonally-located peak values)」在視訊畫面頻譜中的方式，使每個視訊中的畫面皆可藏入一個二位元訊號。而為解決所翻拍的視訊無法與原有影片有相同起始點的問題，以及視訊有許多不同格式的問題，本研究使用了特殊設計的起始及結束訊號及影片格式特性，讓訊息嵌入程序與抽取程序間達到同步。如此一來，使用不同的行動裝置以及不同的視訊格式，都能正確的將訊息嵌入以及抽取。 在訊息抽取程序中，本論文提出了一新訊號偵測的方法，設計了兩個搜尋區域來偵測訊號，且使用了兩個特殊門檻來決定訊號樣式。接著，本論文使用了一線性資料結構，用以重現被嵌入於視訊中的訊息，此結構可使訊號更易於分析與轉換，使得所提抽取程序可成功的抽取出被嵌入的訊號。最後，本論文使用視訊取樣之方式，來加速訊息抽取程序，來提高所提方法的實用性。 深入的理論分析及實驗結果顯示該系統具有可行性及實用性。

In our lives around, many kinds of identities existing in the environment can be made use of to accommodate information for the purpose of pervasive communication. These identities may all be regarded as multimedia in a broad sense. They may appear in many forms like hard copies, advertisements, posters, displays on monitors or TVs, etc. Messages may be injected into such identities by data hiding techniques. These messages can be extracted by using mobile devices such as smart phones, tablets. Such a type of multimedia embedded with messages is called message-rich multimedia. In this study we have proposed a new technique for creating message-rich videos, by which, differently from traditional ways of information transmission, people can exchange message without sending directly digital files such as videos and images. Furthermore, the quality of the newly-generated message-rich video is less-distorted than those proposed by a previous method [26]. Moreover, the message embedded in the video can be extracted in a display-and-imaging fashion. To achieve this purpose, a message embedding process and a corresponding message extraction process have been proposed. For the message embedding process, the use of diagonal peak values in the frequency domain is proposed, by which the proposed message embedding method can embed a 2-bit signal in each video frame. Next, to solve the difficult issues of recording the displayed video exactly from its starting point as well as dealing with multiple video formats, the proposed embedding method makes use of specially-designed initial and ending signals as well as certain properties of the video frame rate to achieve signal synchronization in the message extraction process so that the messages can be embedded and extracted correctly using different mobile devices with different video formats and rates. For the message extraction process, a new signal detection method has been proposed, which is based on the use of a more precise technique of setting two search regions to detect the peak values and using two thresholds to determine the signal patterns. Also, a linear structure is utilized to represent message signals by which the signals can be extracted, analyzed, and converted into ASCII codes effectively. Furthermore, a scheme based on video down-sampling has also been proposed to speed up the message extraction process effectively. The feasibility and effectiveness of all the proposed methods have been demonstrated by theoretical analyses and good experimental results.