動畫轉換為漫畫之自動化系統

Abstract

現今有許多精彩有趣的故事劇情，被製作成動畫的形式供人們觀賞，雖然動畫提供了完整的聲光體驗，但是當我們觀賞動畫時，往往需要耗費許多時間，如果將相同的故事劇情以漫畫的形式呈現，我們可以依照自己的節奏來瀏覽，以較少的時間了解相同的故事內容。然而漫畫的製作過程相當繁雜，需要非常多的人力介入，因此，本論文研製如何以自動化的方式，將一部動畫影片轉換成漫畫。 我們提出了一個針對動畫影片轉換成漫畫的系統，先以影片中連續兩張畫面的差異作鏡頭邊界的偵測，再配合字幕檔案，將影片的鏡頭分類成有字幕的鏡頭以及無字幕的鏡頭，對不同類型的鏡頭分別以不同的方式作關鍵畫面擷取，最後將擷取出的關鍵畫面配置到漫畫畫格上得到一部漫畫。

Nowadays there are many astonishing stories which have been made into animation for people to watch. Although an animation provides a complete experience with visual and sound effect. But when we watch an animation, it takes a lot of time. If the same story is presented in comic form, we can browse it according to our reading pace, spend less time to understand the same story content. However, the creation of comics is quite complicate, it needs a lot of human work involved. Therefore, developing an automatic conversion system for transforming an animation video into comic is necessary. In the thesis, an automatic system to convert animation videos into comics is proposed. First, the difference between two consecutive frames is used to detect shot boundary of video. Then with help of subtitle file, shots are classified into two categories, shots with subtitle and shots without subtitle. For the different category of shots, we provide different approaches to extract the representative key-frames. Finally we map the key-frames onto the panels in comic layouts to form a comic.