Purchase this article with an account.
Peter N. Steinmetz, Flavio DaSilva; Categorizing blurred images. Journal of Vision 2006;6(6):275. doi: 10.1167/6.6.275.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Previous studies have examined how well individual images of faces can be recognized even after blurring (Gold, Bennett and Sekuler 1999, Kornowski and Petersik 2003). We were interested in how well normal subjects can categorize images drawn from several picture categories (unfamiliar faces, animals, tools, outdoor scenes, buildings), both when blurred and unblurred, since this task with unblurred images has been used to investigate single neuron responses in the human medial temporal lobe (Kreiman et. al 2001). Subjects were shown images on a computer screen for one second. Each image was unblurred or blurred by low pass filtering with cutoffs of 0.39, 0.59, 0.78, 1.17 cy/fw. Subjects were asked to indicate, as quickly as possible, whether the image belonged to a target category or not by pressing a mouse button with either the right or left hand. We examined the accuracy of categorization for each image category as well as response times as a function of blur level. Preliminary results suggest that categorization of images is much more tolerant of image blur than image recognition: 75% correct for 0.78 cy/fw (categorization) vs. 8 - 16 cy/fw (recognition). Additionally, the categorization of faces is consistently 54 ± 8 ms faster than the categorization of non-faces for all blur levels. These results are consistent with the existence of a separate form of rapid low frequency sub-cortical processing of faces.
This PDF is available to Subscribers Only