comes from studies not, not exclusively, but comes largely from studies of Random Dot Stereograms, and what's called the Correspondence Problem. A problem that's made [COUGH] especially evident in, Random Dot Stereograms. You've probably seen Random dot stereograms and somewhat different but popular, dot stereograms that, I don't see them too much these days, but not too many years ago you'd find them in the Sunday paper. Because they're using find the stereoscopic object in an autostereogram. Random dot stereograms were discovered in the early 1950s and they were discovered not, not by Bela Julesz, the person I'm going to talk about here. But, by another vision scientist named Ashen Bredder, and the reason they were discovered really had to do with military applications. But, Random dot stereograms, which I'll explain to you in a second aid in the breaking of camouflage and that obviously has. Military interests. The contribution of Julesz, who was a very interesting guy. Was to figure out how to make these Random dot stereograms digitally, on computer technology that was just coming into general application. In the late 1950s, when he began this work. But Julesz was trained in Hungary. He got his PhD there, and he emigrated to the United States in 1956, at a time when Hungary was, you may recall, under very enormous political upheaval, and He was a young man then and, and got his first job at Belabs. You remember we talked about the Bel laboratories before in the context of information through Shannon and enormous contributions of Belabs made to many aspects of science and visual science through the years. And as I said, what Julesz did was to figure out how to make Random dot stereograms easily on a computer. So what's a Random dot stereogram? So this is the left eye view. And the right eye view. And if you fuse these two images. And many of you will be able to do this by what's called free fusion, just by
looking through the plane of the computer
screen, or looking in front of it, to put these two things together in your perceptual minds eye so to speak. And you'll be able to see that, an object in depth pops out of that. So, what a stereogram is, is basically completely camouflaging any object that's in the scene, save an object that is, different in the right and the left eye view, by virtue of some random dots, that don't represent an object unless you fuse these images stereoscopically. So, the great contribution of Random dot stereograms. Basically to get rid of the idea of seeing objects in stereo depth and demonstrate that you can see them in perfectly camouflaged scenes if you simply shift a set of Random dots. So the way of setting this up is indicated in these. Upper diagrams you select in this eye view, just to same Random dots that you're going to shift in the view of your other eye, you've shifted by tiny amounts, some background dots are covered. You fill in the gap that's left on this side with some more Random dots, and, you've generated. An image that's perfectly camouflaged in the sense that if you go to that if you had a steroscophany there's no image there, you can't see you can't see the shift in dots unless you fuse the images and bring stereopsis to bare. And when you do that as I said you see an object. A square that's either in front or in back of the plane of the rest of the random dots depending on whether your fusion is crossed or uncrossed. There are a variety of ways of doing this. If you don't know how to do this, in free fusion, you can. Use a piece of cardboard to help you out and put it between the two eyes, and look down on these and that's helpful. If that doesn't for you, you can buy for a few dollars a little stereoscopic viewer on the internet and look at any stereoscopic image. That way, which is fun to do, I mean, the reason that the autostereograms and stereograms are popular is that it's really fun to look at them. And, of course, as I've said before, this is the technology, so to speak, not much of a technology
that lies behind the.
Viewmaster stereoscope that Wheatstone invented and that you can now buy. So that's great. Where the problem comes in, and where Random dot stereograms have had an impact is making abundantly clear. The so called correspondence problem. And this is in a sense the manifestation of the Inverse Problem, in, in stereopsis. So, let's go back to the proceeding slide. Imagine the difficulty in principle. The brain actually accomplishing what's going on here. Because in some sense, one has to match one's brain. One's visual brain has to match Random dots in the left eye view with the right eye view. And how is that possible? I mean, that seems [COUGH] either computationally or conceptually, just common sensically. The possible task, and everyone who's played with Random dot stereograms is well aware of that. And the name the correspondence problem, has been attached to this conception of difficulty of, how can you find the dot in the left eye that corresponds with the image dot. The right eye to put together these images and see stereoscopically. So as I said, that's a manifestation of the inherent ambiguity in stereopsis that we've talked about in the context of. Basically, all the other aspects of vision, that we've discussed in this series, and just to remind you this is the image that I've shown you a couple of times now. Where the same image on the retina, is generated by objects of different sizes, different distances. And that conflation. Of [COUGH] information in the physical world conflation at the level of the retinal projection is abundantly clear and dramatically clear in the case of Random dot stereograms for the ambiguity that's diagrammed here is. It seems overwhelming how, how to think about how the visual brain could. Overcome that problem, match Random dot patterns in the left, and the right eye. So, this raises the question, and a lot of people are thinking about this, today. Could there be a way around the correspondence problem, and
a simpler explanation of stereopsis?
That is to say the correspondence problem really seems an obstacle that so far hasn't been surmounted in the sense that nobody's really explained how that, problem could be solved. There are a lot of theories that involve complicated computations of, of phase information, energy information, in the retinal image. And well, maybe that's what the visual brain i, is doing. But it seems an uphill battle to think of the visual brain doing that. And one wants to think, or one's really impelled to think about, could there be another way. And the answer is, one really doesn't know but at least I'll want to. I present to you, another possibility that's well worth thinking about. So in this image, what's being compared are correspondence. So this is the world out here. And in this world, here is the horopter that we talked about. And here's a, a little object in the world that's indicated as this brown square, and if you look along to the left, right, left, and the right retina, the, these are the corresponding images of the brown object in, in space. And as I pointed out to you before, the position on the retina of the round projection from this object of the left eye view is, of course, in a different position, given that this object is further than the viropter. In the left eye view or the right eye view, they occupy different positions, and that's where the correspondence problem comes from. How do you match, those corresponding regions? Well, there's another aspect of a projection of images, and that's shown in this kind of light tan. Overlay here, this has to do with corresponding receptive fields. So, corresponding receptive fields depend on retinotopy on the anatomical layout, retinotopically in the left eye and the right eye. And, if you consider projections around the corresponding. Receptive fields, then that's built into the system. There's no correspondence problem in the sense that anatomy has built in, the match in the left eye and right eye
receptive fields at a given point
in retinal on the on the space of the retinal a surface. So, one thing that people are considering these days. Is does that present another way of thinking about stereopsis. That would depend on the relative activity, of the anatomically corresponding Left eye and right eye receptive fields. A way of getting around the correspondence problem and, and, this is simply where this field is at today. It's not clear that that's, going to work. But, it is clear that the correspondence problem, is really a fundamental obstacle in thinking about, the mechanism, the physical way or the. The logical way that stereopsis is generated in the visual brain.