色彩在對稱知覺各階層的角色

Abstract

對稱是一高階的視覺特徵，其辨識須仰賴人類視覺系統的複雜運算。在自然情境中，對稱圖形或物體通常伴隨著多種顏色，因此視覺系統必須整合顏色和形狀的訊息才能偵測到彩色的對稱圖形。本研究以五個實驗來探討顏色在對稱知覺中所扮演的角色。我們將對稱偵測機制分為兩階段—配對與統合，並分別檢驗此兩階段是否對顏色有選擇性反應。研究結果顯示此兩階段均對顏色有選擇性的反應，此表示視覺系統中有一組對顏色選擇性反應的對稱偵測系統。我們進一步操弄了影像中所包含的顏色數目，來探討視覺系統如何整合這些對稱偵測系統來偵測彩色對稱。研究結果顯示無論觀察者是否知道對稱圖案的對稱軸方向，一影像中顏色數目的增加都可以促進對稱偵測的表現。我們也比較了當兩不同顏色對稱圖形共享相同的對稱軸與否時的對稱偵測表現，以探討是否不同顏色能區隔兩圖形而促進對稱的偵測。研究結果顯示，當兩圖形共享一對稱軸時，其對稱的偵測會比兩圖形的對稱軸方向不同時要佳。我們提出一整合了線性的對稱偵測機制、非線性的反應模式、反應干擾的特性與多個系統的決策歷程的彩色對稱偵測模型來解釋以上的結果。此模型顯示一影像中顏色數目的增加會減少對稱偵測系統中的抑制作用，進一步促進對稱偵稱的表現。此外，當兩對稱圖形共享同一對稱軸時，兩對稱偵測系統間的抑制作用會減少，而導致其對稱偵測表現較兩圖形的對稱軸有不同方向的對稱軸時來得佳。

Symmetry is a higher-order form that requires a complicated computation in the visual system. In a nature scene, symmetric objects or stimuli may come with any combination of color. Hence, human visual system needs to integrate both color and form information to detect chromatic symmetry. In this study, we conducted five experiments to investigate the role of color in symmetry detection. We distinguished two stages, that is, matching and pooling stage, of the symmetry encoder and examined the color selectivity of these two stages of symmetry encoder. Our results showed that these two stages are color-selective. This suggests that there are a band of color-selective symmetry channels in our visual system. We further manipulated the number of the colors in the images to investigate how human visual system integrates the response of these symmetry channels to detect chromatic symmetry. Our results showed that the increment of the number of the colors facilitated the symmetry detection performance, regardless the observers had prior knowledge of the symmetry axis orientation or not. Finally, we examined the symmetry detection in two images sharing the same axis or not, to see whether the segmentation of two images with different colors helps symmetry detection. The results however showed better symmetry detection performance when two symmetric patterns shared the same axis than those did not. All these results can be accounted for by a computational model that incorporated linear symmetry encoding mechanisms, nonlinear transducer response, noise manipulation and a multiple channel based decision making process. The model fitting results suggests that the increment of the number of the color reduces the inhibition of the symmetry channels, and in turn facilitates the symmetry detection performance when the images contain more than one color. In addition, the inhibition between channels responding to the two symmetric patterns sharing the same axis is smaller than that between channels responding to two patterns in different axes, in turn facilitates the symmetry detection performance when the two images shared the same symmetry axis.