由影片自動生成「Spot the Differences」遊戲

Abstract

本論文提出一個方法，可利用一段固定視角的影片產生一組「Spot the Differences」遊戲，而我們也是第一個提出自動產生此遊戲的概念。所謂的「Spot the Differences」，也就是一般俗稱「大家來找碴」，是一種以兩張相似的圖片為組合的謎題，遊戲目的是玩家必須找出其中所有的相異之處，然而「Spot the Differences」這遊戲最令人感到有趣的地方在於，玩家有時會對劇烈的變化視而不見，這種現象便是所謂的「變盲」（Change blindness）。我們提出一個有效的方法，除了能有效提高變盲發生率，並且也以物體為基礎定義了「物體視覺注意程度」以及「物體易發意識程度」，前者提供了選擇物體當作相異處的準則，後者則判斷了擺放位置的優劣，而我們便是以這三項作為製作遊戲的基準。最後透過實驗的結果佐證了這三項對於人眼搜尋及辨識相異處的重要性。

In this thesis, we propose an approach to generate a “Spot the Differences” game from a video automatically, and it is the first time that this kind of idea be introduced in computer science. “Spot the differences” is a name given to a puzzle where two similar image create from on image are shown side by side, and the player need to find all the differences between them. The most interesting part of this puzzle game is that players sometimes ignore the obvious difference while playing. This phenomenon is called “change blindness”. In order to generate a challenging “Spot the Differences” game, we propose an approach not only increasing the occurrence rate of change blindness but also define the two important terms in this thesis-“the attention rate of object, called ATL” and “the awareness rate of object, called AWL.” Based on the definition of ATL and AWL, we choose which object to be difference and decide the position where to place. In the last, we design two experiments to prove the correctness’ of our assumptions and approach.