Technology to increase digital resolution by factor of three for sensor?

[henry_minsky1]

I was thinking, since the Canon digital sensor uses a color filter

over the pixels, what if the color filter were on a layer that could

slide, and a piezoelectric element shifted the filter one pixel up

and then one pixel over, taking three images in rapid succession,

then you would have three times the resolution, i.e., a real RGB for

each pixel like the foveon gets. The trick would be clearing the

sensor data quickly enough to take the different pictures fast

enough to get a decent equivalent shutter speed. Or is the

mechanical shutter too slow for this? I wonder what the speed

bottleneck is, the mechanical shutter or the electronics to reset

the pixel sensor elements?

[wedding-photography-denver]

Wow, what a fun and light hearted subject.

 

Why would this be of use since the Foeveon is already doing true RGB? Could you not just increse the physical size of that sensor?

 

Regards.

[robert_hively]

Maybe the next 1Ds mark 2 will be the first three sensor camera?

 

http://www.pcmag.com/article2/0,1759,1606046,00.asp

[gary_ferguson1]

Canon patented this a couple of years ago, it was recently pointed out on another web site.

[omeryair]

This idea is partialy used by some medium format digital backs: the sensor reads the data, then shifts half a pixel and records more data. This doubles the sensors outpus.

[ron c sunshine coast,qld,a]

In practice you can't get three times the resolution because the current interpolation used with bayer sensors is really quite good.

<BR>Notice the bayer sensor isn't in practice three times better per megapixel-it's only about 50%(?) depending on who you ask

[mark u]

Maybe you should look at some of NASA's interpolation algorithms for larger pixel monochrome (or multiple exposure multi filter) capture that allow sub pixel size resolution to be extracted from images. The trick to apparently exceeding Nyquist limits is having some knowledge of what is being imaged. The basic idea is that if pixel A shows red, and pixel B a shade of purple, while pixel C shows blue, then you can estimate a sharp boundary position between red and blue within the width of pixel B according to the shade of purple. You can't really use these techniques with a Bayer sensor though - where pixel B could be green and show no response to either blue or red, for example.