ISSN : 1226-9654
The previous models of human symmetry perception can be summarized as a point-by-point comparison model (Julesz, 1971) and a density comparison model (Barlow & Reeves, 1979), both of which are single-channel model based on spatial domain. Neither of these spatial domain model, however, cannot account for the relative effectiveness of symmetric information as a function of the distance of the information from the axis of symmetry. In order to incorporate the relative effectiveness from the axis on symmetry perception into the density comparison model, the study by Barlow & Reeves was replicated and the weights for each spatial location were calculated by multiple regression. By this way, the degree of symmetry of an input image is a weighted sum of small areas, not a straight forward sum. In experiment 2, however, in which no dots in the areas near the axis was presented, the weighted density comparison model fails to account for the human observers' performance. With the other research results showing that the detection of symmetry as a combination of two orthogonal symmetries depended on the two independent symmetries (Julesz & Chang, 1979) the present result suggests that human symmetry perception may be processed through multiple channels, which differing in resolution and/or orientation.