A Study of Content Analysis Based Virtual Product Placement

由於數位多媒體內容爆炸性的增長和無遠弗屆的散播能力，將廣告訊息搭載多媒體內容而傳播的廣告方式，創造了後勁十足的經濟效益。隨著以內容為基礎的多媒體分析和電腦動畫方面之技術的進步，虛擬置入式行銷引領出未來的新趨勢，並且開展了更多的機會，能以比較有效率的方式來為多媒體資產創造更多附加的經濟價值。許多重要而且挑戰的議題因此相應而生，例如，如何將與多媒體內容或是觀看群眾在語意上相關的廣告，以一種令人印象深刻卻比較不造成侵入式干擾的呈現方式，有效率地在適當的時間點置入於多媒體中合適的區域上， 使得廣告訊息的傳達效益能夠最大化。 在本論文中，我們提出一種一般化的架構和相對應的設計理念，以處理所面臨到的關鍵性問題，並且將虛擬置入式行銷的置入型態，由單純的螢幕置入推向至更有廣告效益的情節置入。為此，我們同時考慮廣告生態系統中各個角色的本質與特性，利用了以內容為基礎的多媒體分析技術，以及詳盡的相關領域知識，讓虛擬置入式行銷能更有廣告效益，並且減少侵入式干擾。基於所提出的一般化架構，我們針對數位多媒體中的影片類型，實現了示範性的虛擬置入式行銷系統，其中發展並且探討了兩種以虛擬地視覺上產生互動為基礎的廣告置入機制：1) 廣告外觀隨時間變化地置入於空間上固定的區域，2) 廣告外觀隨時間變化地置入於空間上不固定的區域。在第一種廣告置入機制當中，廣告會被置入於影片裡自動被偵測到的合適時空，即當觀眾會注意影片中吸引人的物件時，場景內統計上有相對較高的機率被注意到的區域。接著，根據目標影片的內容，將置入的廣告在外觀上作動態的調整，使得廣告訊息可以透過潛在的通訊管道—也就是人眼的餘光，以比較不造成侵入式干擾的方式傳送給觀眾。至於第二種的廣告置入機制中，要置入於影片當中的廣告，會根據預先定義的行為模式，以一種會隨著該影片內容而演變的動畫效果呈現於其中。如此一來，被置入的廣告可以與影片中的內容物產生互動並且演化，因而引發附加的故事支線以進一步給予觀眾較深的印象。 本研究代表了第一個從互動性的視覺感官感受的角度，嘗試實現情境置入的效果，並且透過以內容為基礎的多媒體分析技術的輔助，展示了虛擬置入式行銷的可行性和有效性The explosive growth and widespread distribution of digital multimedia contents creates huge potential revenue in taking multimedia contents as information carriers for advertising. With the advance of techniques in computer graphics and content-based multimedia analysis, virtual product placement signals the new trend and opens up additional opportunities to effectively monetize the multimedia asset in an efficient way. Accordingly, a number of significant and challenging issues are raising, such as how to efficiently insert the semantically relevant advertisements at the appropriate place and time with the impressive yet less-intrusive representation in the multimedia contents, so as to maximize the effectiveness of advertising communication. In this dissertation, a general framework and design philosophy are proposed to tackle the critical issues and extent screen placement toward plot placement for improving advertising effectiveness of virtual product placement. To this end, techniques in content-based multimedia analysis and exhaustive domain knowledge in support of effective and less-intrusive virtual product placement are leveraged while simultaneously considering the nature of each role in an advertising ecosystem. Based on the proposed general framework, we realized exemplary systems dedicated to virtual product placement in videos, where alternative virtually visual interaction (V2I) based ad insertions are developed and investigated: space-still time-varying (SSTV) V2I based ad insertion and space-varying time-varying (SVTV) V2I based ad insertion. In SSTV V2I based schemes, advertisements are inserted in the automatically detected points that are with relatively larger probability of being noticed while watching the attractive object in videos. By dynamically adapting the appearance of inserted advertisements to the source video, advertising message can be less-intrusively delivered to viewers through the latent channel: the extraneous visual acuity of human. As for SVTV V2I based ad insertion, advertisements are inserted into videos with evolved animations according to predefined behaviors, so that the inserted advertisements can interact with contents in videos and induce additional storyline to impress the viewers. The studies represent the first attempts toward plot placement from the perspective on visual perception of interactivity and demonstrate the feasibility and effectiveness of virtual product placement with the aid of content-based multimedia analysis