Purchase this article with an account.
Tatsuto Takeuchi, Karen K. De Valois; Motion sharpening in moving natural images. Journal of Vision 2002;2(7):377. doi: 10.1167/2.7.377.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Moving objects often appear normal or even sharper than they actually are. This effect, called motion sharpening, has been demonstrated empirically. Video-taped images are actually somewhat blurred, but they look sharp when the video is played. Most studies regarding the motion sharpening have used simple stimuli, such as moving gratings. We have examined this sharpening process by varying the amount of blur in moving natural images.
We prepared movies from natural images (running animals, flying birds, aerial landscapes, river flow, moving CG images, and so on), each of which comprised 100 frames. X of the 100 frames were blurred by low-pass filtering and combined with the non-blurred frame to make a movie of 30 fps. Subjects judged the perceived sharpness of the movies by comparing the movie to a blurred still image.
When the amount of blur was small (1/20 of the compression rate in the JPEG algorithm), motion sharpening was perceived even when all 100 frames were blurred. This was shown earlier. However, no sharpening was perceived if all 100 frames were blurred to a level at which the details of the images were disappeared (1/200 of the compression rate, in which the spots on the body of a cheetah disappeared, for example). However, we found dramatic motion sharpening even when 70 frames were greatly blurred (so that the movie contained only 30 sharp image frames). Even with 90 frames greatly blurred, the appearance of the movie was still acceptable, though some flicker between blurred and non-blurred images was perceived.
Since the amount of blur is very large in the above case, our results suggest that motion sharpening is stronger than previously recognized. This kind of strong motion sharpening depends on the spatial structure (layout) of the images. The sharpening is stronger when the background pattern is moving (in the situation when a moving object is tracked by a camera, for example), which suggests contrast gain control in a motion detecting mechanism.
This PDF is available to Subscribers Only