Modelling Responses to Uncomfortable Images

L O'Hare1, A Clarke2, P B Hibbard3

1School of Psychology, University of Lincoln, United Kingdom
2School of Informatics, University of Edinburgh, United Kingdom
3Department of Psychology, University of Essex, United Kingdom


The visual system is thought to be optimised to encode typical natural images, to keep the metabolic costs of processing low (Barlow, 1961, in: Sensory Communication, W. A. Rosenblith, MIT Press). Field (1994, Neural Computation, 6, 559-601) showed that natural images are efficiently (sparsely) coded by a model visual system consisting of wavelet filters, based on known properties of cells. Images deviating from the statistics of natural images, e.g. stripes, have been shown to cause viewing discomfort, such as headaches, eyestrain, and distortions of vision (Wilkins et al, 1984, Brain, 989-1017). Juricevic et al (2010, Perception, 884-899) suggested that discomfort could occur when images create an excessive neural response. The current study used a simple, physiologically based model of V1 to assess the sparseness of the response to uncomfortable stimuli. The kurtosis of the population response to uncomfortable images was lower than to natural images, and showed the same spatial frequency tuning as discomfort judgements. This suggests that the population response to uncomfortable images is less sparse compared to natural images, supporting the suggestion that discomfort can be caused by excessive metabolic demands.

Up Home