透視一致性之敏感度

Abstract

外在環境是具有深度的三維空間，我們依賴各種不同來源的線索建構大腦內的立體影像，而人類從古至今也不斷的研究各種方法，探索如何在二維的平面架構自然寫實三維空間實景。透視法的基本原理在十五世紀即被研究透徹，由希臘羅馬壁畫中的場景到文藝復興時代的數學幾何方法，數學幾何又直迄今日攝影法與3D軟體之發展。其發展雖是文藝復興時期西方繪畫的最重大成就，但在大部分重要畫家的作品中均不難找到局部違反透視原則的錯誤。 因此本研究欲針對人對於透視的不一致性之敏感度進行測試，於實驗中製作一點、二點與三點透視立方塊圖形，以及改變觀察軸及移動觀察點等刻意造成畫面中之透視錯誤，做出物件消失點不一致之刺激圖形。消失點變動的方式為水平方向的移動。本實驗要求受試者於一個具有三到四個立方塊物件的畫面中，指出那個方塊的透視點與其餘方塊不同，並利用簡單迴歸進行分析。 研究結果指出（1）消失點的變動是在水平方向移動，則受試者對於此偵錯作業的表現都非常不理想，僅在三點透視的情況能夠勉強高於猜測的答對水準，一點與二點兩點透視的情況與猜測機率無異。（2）消失點移動的絕對量對於偵錯作業的難易程度之預測力薄弱。 本研究的結果顯示人對於透視錯誤，特別是多元消失點在水平方向共線的類型錯誤非常不敏感。此外，本研究亦針對幾點透視、軸的改變與移動單位作為因子進行三因子與雙因子變異數分析，以理解變數間對透視敏感度的影響關係，分析得知畫面中消失點不一致的線索越多及錯誤消失點與原消失點的距離差距越大，受測者更能觀察的出來錯誤。 本研究結果可加深我們對深度知覺的瞭解也可提供未來在藝術與設計的透視學的應用參考。

Although the principles of linear perspective have been understood by artists since the 15th century, inconsistencies in perspective can be found in paintings produced since then. However, most perspective errors go unnoticed without close scrutiny. We propose a systematic evaluation of our sensitivity to errors in perspective in a picture. The stimuli were pictures of blocks conforming to one of the perspective rules (single-point, two-point, or three-point perspective). The participant had to pick out a target block that did not agree with the others in perspective. The picture of the target block was taken at a different shooting angle from that of the other blocks. The camera settings for shooting the target were manipulated by moving the camera (i) laterally, (ii) forwards/backwards, and (iii) in rotation. We found that the ease of error detection increased with conditions in the above order. The performance in the single-point perspective condition was only slightly better than chance. The correlation between the absolute displacement of the vanishing points and the performance was evident only within condition. The results indicate that the type of error, or the number of available cues are more important than the magnitude of error in accounting for error detection performance.