Full Frame Sensor - One large pixel

drlesser

New member
Messages
5
Reaction score
0
Location
MD, US
If it was possible to construct a full-frame (35 mm) sensor that was one continuous "crystal" (ie., 1 very large pixel), would the image quality be superior?
 
Imagine that it rains in red green and blue rain drops. Now imagine that you put out 12 million rain guages in your back yard. Each rain guage has a trick filter on it that only lets one color of rain drops into that particular guage, and is marked off with 255 marks, you also put out twice as many green rain guages but that's not really important right now.

So it comes a rain, you flip a switch and the lid to the rain guages open, and you keep it open for a while. Then you flip the switch again and all the lids on the rain guages close. Then you go out and count the number on each of the 12 million rain guages and write down exactly where it was.

That's what your camera does now.

Now what you're saying is lets just dig up the back yard and put in a big swimming pool. Then when it rains we'll take the cover off the pool, let it rain for an hour, cover it up, then go check and see how much it rained when it's over.
That's what your one pixel camera would be like. See the difference?
Probably not, I probably just made it even more confusing.
 
the dynamic range of a one pixel image could always be interpreted as perfect.

no blown highlights, no blocked shadows teh full range of tonality is represented properly in the digital capture. of course there is only one tonality value to capture ..

kind of a silly discussion too ;-)

--
Michael Salzlechner
http://www.PalmsWestPhoto.com
 
Not really a silly discussion. I think I understand much better the concept of pixels and pixel capture much netter.
 
Since your question was already addressed here, I won't do it again.

However, the method most current sensors use -- a color filter array in a rectangular matrix -- has a good many downsides. It might be interesting to think of alternatives.

There's Foveon, of course, with the sensels layered on top of each other. The pro is much better per-pixel resolution, but the con is different degrees of sensitivity for different colors, more complex color processing and lower color accuracy, and scattering of light as it penetrates the silicon (i.e., hard limits to the maximum attainable resolutuion).

There are other possible ways of getting true-color pixels too. For example, suppose you could layer something like an LCD on top of the sensor, where you could change its color extremely quickly, and you had a sensor that you can address equally quickly. If you switched between R, G, and B at, say, 12 kHz, you would get a true-color capture with a minimum shutter speed of 1/4000 s. Can't be done with current technology, but it's certainly possible. If you cycled too slowly, though, you'd get temporal aliasing -- moving objects in the picture would get a smear that changed color.

Then, you could use a pseudo-random sensor matrix, with sensels of different sizes. Small sensels would provide resolution but not do great in the shadows, and big sensels would provide tonal information and see great in the shadows. Because the matrix would be pseudo-random, moiré would pretty much go away: no more need for an anti-aliasing filter, and you could leverage the entire resolution of the sensor. In fact, this would closely replicate the imaging characteristics of film -- film is, after all, a random matrix of sensels -- only the sensels are chemical rather than digital. The difficulty is that it would be considerably more complex to demosaic this kind of picture than the one from a rectangular one.

Other ideas you've thought of or come across?

Petteri
--
My flickr page: [ http://www.flickr.com/photos/primejunta/ ]
Me on photography: [ http://194.100.88.243/petteri/ ]
Me on politics: [ http://p-on-p.blogspot.com/ ]
 

Keyboard shortcuts

Back
Top