The effect of sharpness constancy on the recognition of facial expression
Motoyasu Honma, Y. Osada · Journal of Vision · 2010
Video-sequences appear sharpness even blurred images inserted into them. This effect, called sharpness constancy (Ramachandran et al, 1974), has been demonstrated empirically. Video-sequence images are actually somewhat blurred, but they look sharp when the video is played. We have examined effects of motion information of a face on the recognition by the method of varying the degree of blur among facial expressions. Methods: We prepared video-sequences of 2 facial expressions (happy and sad), each of which comprised 26 frames. A video-sequence of 29.97 fps was mixed alternately the half of the 26 frames was blurred by Gaussian filtering with the non-blurred frame. Video-sequences were set by 3 scales (Gaussian filter radius 0, 4, 8 pixel). Also, Blurred still images were set by ten scales (Gaussian filter radius 0–9 pixel) as comparison stimulus. 2 observers judged the perceived sharpness of the movies by comparing the movie to a blurred still image. Results: In case that the value of blur of video-sequences was large (4 or 8 pixel), observers judged the value of blur of still image lower than the value of blur of video-sequences. However, appearance keeps sharpness when video-sequences of sad faces show than when that of happy faces show. Conclusions: We found that sharpness constancy occurs in video-sequences of facial expression, and this effect differs among facial expressions. These results suggest that a motion detecting mechanism on the recognition of facial expression depends on the spatial frequency component of facial expression. Two explanation of this effect are possible: 1. Motion information of a face reconstructs the high spatial frequency information. The reconstruction ratio on happy faces is larger than that on sad faces. 2. Blurred images of a video-sequence were neglected. We can easy recognize the happy face with low spatial frequency, but we are hard to recognize sad faces with low spatial frequency.