Purchase this article with an account.
Michele Rucci, Antonino Casile; Fixational instability and natural scene representation. Journal of Vision 2005;5(8):374. doi: 10.1167/5.8.374.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Images of natural scenes tend to vary smoothly in space and time. It is a long-standing proposal that an important function of the early stages of the visual system is to minimize this input redundancy to allow efficient visual representations. In particular, it has been observed that the response characteristics of neurons in the retina and LGN may attenuate the broad correlations that characterize natural scenes by processing input spatial frequencies in a way that counter-balance the power-law spectrum of natural images. Here, we extend this hypothesis by proposing that the movement performed by the observer during the acquisition of visual information also contributes to this goal. During natural viewing, the projection of the stimulus on the retina is in constant motion, as small movements of the eye, head, and body prevent maintenance of a steady direction of gaze. To investigate the possible influence of a constantly moving retinal image on the neural coding of visual information, we have analyzed the statistics of retinal input when images of natural scenes were scanned in a way that replicated the physiological instability of visual fixation. We show that during visual fixation the second-order statistics of input signals consist of two components: a first element that corresponds to the broad correlations of natural scenes, and a second component, produced by fixational instability, that is spatially decorrelated. This second component strongly influences neural activity in a model of the LGN. It decorrelates cell responses even if the contrast sensitivity functions of simulated cells are not perfectly tuned to counter-balance the power-law spectrum of natural images. The results of this study suggest that fixational instability contributes to establishing efficient representations of natural stimuli.
This PDF is available to Subscribers Only