Purchase this article with an account.
Isamu Motoyoshi, Shiori Mori; Image statistics and the affective responses to visual surfaces . Journal of Vision 2016;16(12):645. doi: https://doi.org/10.1167/16.12.645.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Humans are not only able to perceive physical properties and material of natural surfaces, but also judge their emotional values, which would afford attraction toward, or aversion from, particular surfaces. To investigate visual process underlying such emotional responses, we asked human observers to assess comfortableness and unpleasantness for 193 images of natural surfaces that included clothes, foods, skin, stones, liquid, trypophobic stuff, and so on. We analyzed the relationships between the rating data and image statistics of these stimuli in the cone-contrast space (Lum, RG, and YB channels). The analysis revealed that, for all color channels, the SD at mid-spatial frequency bands and the cross-orientation energy correlation at high-spatial frequency bands were correlated positively with unpleasantness and negatively with comfortableness, respectively (p< 0.01). It was also found that comfortableness was negatively related with the luminance vs. color correlation at high-spatial frequency bands (p< 0.01). We obtained similar patterns of the rating data when we used PS texture images synthesized from the original images (r = 0.76 for comfortableness and r = 0.77 for unpleasantness, respectively) or phase-scrambled images (r = 0.68 and r = 0.69), and when we presented stimuli for a short duration (50 ms) so that observers could hardly recognize the surface category (r = 0.77 and r = 0.73). These results indicate that human emotional responses to natural surfaces largely, not entirely though, depends upon relatively low level image statistics. We suggest visual mechanisms that may utilize textural information to directly summon emotional reactions for surfaces preceding the recognition of their material category.
Meeting abstract presented at VSS 2016
This PDF is available to Subscribers Only