Keywords
Spatial vision, Image gradients
Abstract
How do we achieve a sense of spatial dimension from a sense of location? There are three predominant ideas about how we achieve this; spatial isomorphism, in which what we see reflects differences in distance or size in the brain; that spatial extent depends upon motor sensations or intentions related to eye movements; and that distance is computed from the degree of correlation in neural activity between adjacent locations, with distance inversely proportional to the correlation. There are problems with each of these approaches, for example, neural correlation may depend more on image structure than adjacency - consider the case of images containing repeating lines or sine gratings. Here a new computational strategy is outlined and assessed. Human image motion computation naturally relies on a set of linear temporal filters which can be defined in a window of around 100ms. However, gradient motion models can be reformulated to operate on just two frames, with image frame sums and differences taking the place of extended temporal derivative filters. This approach constitutes an image gradient approach to computing retinal disparity. This strategy can itself be re-purposed to compute the separation of points in a single image, rather than computing image displacement in the left and right eye’s images. This new strategy allows us to compute spatial separation, on the basis of a non-spatial measure, the image brightness difference, and a local spatial brightness gradient.
Included in
From Image Gradients to a Perceptual Metric Space
How do we achieve a sense of spatial dimension from a sense of location? There are three predominant ideas about how we achieve this; spatial isomorphism, in which what we see reflects differences in distance or size in the brain; that spatial extent depends upon motor sensations or intentions related to eye movements; and that distance is computed from the degree of correlation in neural activity between adjacent locations, with distance inversely proportional to the correlation. There are problems with each of these approaches, for example, neural correlation may depend more on image structure than adjacency - consider the case of images containing repeating lines or sine gratings. Here a new computational strategy is outlined and assessed. Human image motion computation naturally relies on a set of linear temporal filters which can be defined in a window of around 100ms. However, gradient motion models can be reformulated to operate on just two frames, with image frame sums and differences taking the place of extended temporal derivative filters. This approach constitutes an image gradient approach to computing retinal disparity. This strategy can itself be re-purposed to compute the separation of points in a single image, rather than computing image displacement in the left and right eye’s images. This new strategy allows us to compute spatial separation, on the basis of a non-spatial measure, the image brightness difference, and a local spatial brightness gradient.