A multi-scale model of binocular fusion in the human visual system

We have two eyes but see one world. Although this might seem obvious (there is only one world), there are conditions where we do see double. For example, hold your index finger up a few inches in front of your nose, then fix your gaze on a more distant object straight ahead; your finger will be clearly seen as two, side-by-side. Then, as you switch your gaze to the finger, the two will become one. This is binocular fusion, or 'single-vision', and it is achieved by the brain (the visual cortex, containing thousands of millions of nerve cells) piecing together the information from both eyes. As a result, your ability to detect very faint things, or to see fine details, is better with two eyes than one. Our research aims to study single-vision and double-vision in carefully controlled experiments, and to build a general explanation in the form of a computer model that identifies both the main mechanisms and how they serve to identify basic visual features such as lines and edges that may be the building blocks of perception. How is the binocular combination organized ? Does the visual system use only the combined (left+right eye) information, or does it also make use of separate responses from the left eye and right eye ? Or might it even use the difference between left and right views, as well as the sum ? Our experiments will get at these questions by looking at how well we can detect images that are, in some sense, opposite in the two eyes. This could be a pattern of light and dark stripes, where the light stripes in one eye are superimposed on the dark stripes of the other eye, or the stripes might increase in contrast for one eye while they decrease in contrast for the other. If the brain only had left+right information available, such opposites should cancel. Previous findings suggest that they don't cancel, and so other information must also be used. New experiments will test this further and clarify what the extra information may be. Whether we see things as single or double depends on how far apart the left and right eye's views are and, importantly, on how blurred or sharp they are. Our second set of experiments will carefully measure these effects using images (blurred edges) that have not been used before, but should give cleaner, more general results than previous studies. Vision in one eye can be dramatically suppressed or inhibited by events in the other eye, especially when images in the two eyes are very different. Bright, flashed images in one eye particularly tend to suppress weak, steady images in the other eye. Present evidence, however, suggests that such competition between the eyes is also part of the normal process of binocular combination. We aim to study several aspects of binocular fusion (the perceived position, contrast and blur of edges and lines), and to do so when the left and right images themselves differ in position, contrast, or blur. With these data, assisted by our computer model, we shall work out how the mutual suppression between the eyes, taking place in the brain, shapes the binocular fusion and appearance of visual features. This work will be of direct benefit to other scientists researching in various fields that are concerned with binocular vision: neuroscientists studying the cellular responses of the visual brain; vision scientists and psychologists studying sensation & perception and brain-imaging; computer scientists interested in machine vision and robotics, especially 3-D vision; and optometrists concerned with vision correction and treatment of sight disorders. In optometry, our work may shed new light on 'monovision', where older patients (40+) are often successfully given different contact lenses in the two eyes - one for far, and one for near. By explaining how images with different blurs in the two eyes are combined by the brain, our research could lead to better understanding of monovision and to improvements in diagnosis and prescription.

The visual cortex combines signals from the two eyes resulting in fusion or diplopia, with or without stereo vision, but fusion and its neural basis remain poorly understood. The goal of this project is a more complete functional account of binocular fusion through psychophysical experiments and computational modelling. Modelling: A key objective is to unify our models of multi-scale feature analysis (JoV 2007) and binocular signal processing (JoV 2006), to implement multi-scale spatial vision in the context of binocular combination and interocular suppression. We need to know (i) how multi-scale filtering is organized across monocular & binocular stages of processing, (ii) how responses are combined across the eyes, (iii) how interocular suppression modifies each eye's response & how it is distributed across space & scale, (iv) how decisions about the identity of features are made. Experiments: (1) to address the functional architecture and performance of the binocular system, by finding whether * cancellation is ever possible between opposite images in the two eyes, * observers can distinguish between monocular and binocular inputs, * observers can tell that fused left and right images are different in contrast. (2) to quantify the disparity that separates fusion from diplopia for blurred test edges, over a range of blurs & contrasts. (3) to study the fused appearance of edges or bars that have different positions, contrasts or blurs in the two eyes, so as to shed light on the summation and interocular suppression that shape the appearance of binocular features. These data and model will answer some key questions about: * the necessary conditions for fusion * the cause of double vision at larger disparities * why the disparity range of single-vision is scale-dependent * what entities are combined between the eyes * the inter-play of feature detection and binocular fusion * the role of monocular and binocular mechanisms in fusion and diplopia.