Cognitive Daily reports nearly every day on fascinating peer-reviewed developments in cognition from the most respected scientists in the field.
If the human eye was a digital camera, how many megapixels would it have?
Clarkvision does the calculations.
The answer: 576 megapixels.
Impressive job -- I wish I had thought to do that. Note that their calculations require a bit of fudging: the fovea actually covers just a tiny bit of the visual field; the eye must move from point to point in order to assemble an image this detailed. A digital camera records all the pixels at the same time. For the photographically inclined, the article also goes on to make several other camera/eye calculations.
A separate question: could a 576 megapixel image "fool" your visual system into believing it was seeing the real thing? Assuming one eye was covered and you were not allowed to move, I think it could. But as soon as you viewed the image with both eyes or were allowed to move, then you would be able to detect the fact that the image was flat. Three-dimensional images look different when viewed from different perspectives, but flat images don't.
Comments
The question should be, can we create a 576 megapixel image?
Posted by: Oskar Syahbana | October 22, 2006 1:56 PM
The question should be, can we create a 576 megapixel image?
Someone did: http://www.tawbaware.com/maxlyons/gigapixel.htm
Posted by: oku | October 22, 2006 3:21 PM
A bit disingenious, I think; if you allow the eye to move and assemble an image over time, then of course you can do the same with a camera (and the gigapixel image linked to above is one example).
Posted by: Janne | October 22, 2006 8:56 PM
As digicam freaks will know, there is a lot more to image quality than resolution. THere is also bit depth - and the human visual system clearly has current consumer digicams soundly beat when it comes to that. At issue is the dynamic range of the images. You can get a feeling for how much consumer grade photos lack by looking at pictures from large format cameras, but even then (I think) you're missing the ultra-high dynamic range that is present in real-world imagery.
Interesting analysis though!
Posted by: Chris Chatham | October 22, 2006 9:23 PM
What about data compression during visual perception? I can't recall it, but I guess that is quite impressive too.
Posted by: sparc | October 22, 2006 11:51 PM
Chris has a point. One of the big giveaways in photography is that most photos represent just a tiny slice of the light that would be entering the human eye at the actual scene. There have been attempts in the world of computer graphics to correct for this limitation. High Dynamic Range Imagery, or HDRI, is a special image format that contains data from multiple exposures for one "picture", so that you can get an idea not just of what's bright or dark in a scene, but what's emitting and absorbing the most light. Paul Debevec has been doing research into this for years, with special emphasis on reprojecting the HDRI data for use in special effects and scene simulation. His research has been used in the Matrix movies. See http://www.debevec.org Pay attention to the old, but still very impressive, Fiat Lux, in which Debevec recreates the interior of St. Peter's Basilica using sampled light data.
Posted by: Jonathan Dobres | October 23, 2006 9:16 AM
How the human visual system 'resolves' visual information is only mildly related to how a camera captures a picture. Even at the level of the retina, the amount of computation involved in contrast adaptation is staggering. How visual information is represented in early visual processing is still a mystery. Not even mentioning how feedback such as spatial and global attention might influence individual channels of information.
In the lateral geniculate nucleus, for example, approximately 80% of all neurons representing the 'visual image' are feedback neurons from higher visual centers, where the other 20% are the feedforward neurons from the retina.
Posted by: Alan Horsager | October 23, 2006 11:04 AM
Yeah, 576 million pixels of functional detail from only 6 million cones (the color-detecting photoreceptors) and 125 million rods (mostly for peripheral vision).
On a related note, I find the work to create camera systems for blind people extremely interesting, but very premature. They've created cameras that sit on your glasses and send impressions to your tongue or chest or whatever, but to pan the camera you have to rotate your head. What they really ought to be doing is to have little eyetrackers (perhaps using little magnets in your eye or something easy to track) that can figure out where you're looking and selectively show the fovial equivalent on the tongue/chest. No need to point the camera, just use a fisheye lens and take a different portion of the image. This would probably be more useful to users of this kind of device than increasing the resolution.
Posted by: Harlan | October 23, 2006 11:13 AM
how do u calculate 576 megapixels for eye?
Posted by: VINDISHA | May 7, 2008 12:49 PM