Sigma/Foveon Questions
in [Digital Cameras]
|
Prev: SI Facescape
Next: FF camera with mirrorless design
From: David J Taylor on 28 Apr 2010 04:51 "nospam" <nospam(a)nospam.invalid> wrote in message news:280420100124302092%nospam(a)nospam.invalid... > In article <hr8mv8$tah$1(a)news.eternal-september.org>, David J Taylor > <david-taylor(a)blueyonder.co.uk.invalid> wrote: > >> Consider printing a 3MP image at 10 x 8 inches, for example. You get >> print A. Now upsize the image to 12MP in your image processing >> software, >> and print it out again at 10 x 8 inches. Print B. Prints A and B are >> the >> same size, but the image has been increased in resolution in between. >> So >> an upsized image, but at the same size. "Size" can equally refer to >> the >> number of pixels as to the final print size. > > upsizing doesn't add any detail so the resolution isn't going to be any > different. I think that many people would see the difference between a straight 3MP image and an interpolated 12MP one, in a 10 x 8 inch print. Don't you? Cheers, David
From: David J Taylor on 28 Apr 2010 05:02 "nospam" <nospam(a)nospam.invalid> wrote in message news:280420100132190252%nospam(a)nospam.invalid... > In article <hr8n74$uku$1(a)news.eternal-september.org>, David J Taylor > <david-taylor(a)blueyonder.co.uk.invalid> wrote: > >> One imager I work with has 11 pixels at each spatial coordinate: >> >> http://www.esa.int/esapub/bulletin/bullet111/chapter4_bul111.pdf >> >> and there are 3712 x 3712 coordinates in an image. > > the word pixel appears only once in the entire document and it says > 'image data' for 4 visible/near-infrared and 8 infrared channels, not > pixels. > > it's a 13.7 megapixel sensor (3712 x 3712), with each pixel having 12 > components. it is *not* 165 megapixels. If you look at the physical sensor, there are (IIRC) three detectors for each of 11 channels, so does that make it a 33 pixel sensor? BTW: channel 12 actually has three times the resolution, and has 9 detectors scanned. Where it differs from the Bayer sensor is that at each spatial location, 11 different spectral bands are sensed. so that at each location there are 11 independent measurements. No interpolation involved. With the Bayer sensor there is only one measurement at each location, so in the 12MP sensor, 3MP of red, 3MP of blue and 6MP of green, and yet in the final image there are 12MP of red, 12MP of green and 12MP of blue. These extra values are derived by spatial interpolation and other refinements. The SEVIRI more resembles the Foveon approach, but with far better filtering to separate the individual spectral components. For each channel, 3 detectors, and 3712 x 3712 pixels. Cheers, David
From: Peter on 28 Apr 2010 06:45 "David J Taylor" <david-taylor(a)blueyonder.co.uk.invalid> wrote in message news:hr8mv8$tah$1(a)news.eternal-september.org... > "Ray Fischer" <rfischer(a)sonic.net> wrote in message > news:4bd7d114$0$1667$742ec2ed(a)news.sonic.net... >> Kennedy McEwen <rkm(a)kennedym.demon.co.uk> wrote: > [] >>>No size change need occur when you upsize an image from 3Mp to 12Mp >>>either. >> >> You are an idiot. >> >> -- >> Ray Fischer >> rfischer(a)sonic.net > > Consider printing a 3MP image at 10 x 8 inches, for example. You get > print A. Now upsize the image to 12MP in your image processing software, > and print it out again at 10 x 8 inches. Print B. Prints A and B are the > same size, but the image has been increased in resolution in between. So > an upsized image, but at the same size. "Size" can equally refer to the > number of pixels as to the final print size. You're wasting time trying to explain terms of art. He thinks a table of contents is a piece of furniture. -- Peter
From: nospam on 28 Apr 2010 09:58 In article <hr8td6$u8d$1(a)news.eternal-september.org>, David J Taylor <david-taylor(a)blueyonder.co.uk.invalid> wrote: > >> Consider printing a 3MP image at 10 x 8 inches, for example. You get > >> print A. Now upsize the image to 12MP in your image processing > >> software, > >> and print it out again at 10 x 8 inches. Print B. Prints A and B are > >> the > >> same size, but the image has been increased in resolution in between. > >> So > >> an upsized image, but at the same size. "Size" can equally refer to > >> the > >> number of pixels as to the final print size. > > > > upsizing doesn't add any detail so the resolution isn't going to be any > > different. > > I think that many people would see the difference between a straight 3MP > image and an interpolated 12MP one, in a 10 x 8 inch print. Don't you? no. it might be slightly better than just letting the printer handle it, but do you really think there will be more detail in the upsized one?
From: nospam on 28 Apr 2010 10:09
In article <hr8tk2$vhf$1(a)news.eternal-september.org>, David J Taylor <david-taylor(a)blueyonder.co.uk.invalid> wrote: > If you look at the physical sensor, there are (IIRC) three detectors for > each of 11 channels, so does that make it a 33 pixel sensor? no. fuji had a similar problem. if you had a single dot with a million layers or channels or whatever, would that be a megapixel sensor? > BTW: channel 12 actually has three times the resolution, and has 9 > detectors scanned. > > Where it differs from the Bayer sensor is that at each spatial location, > 11 different spectral bands are sensed. so that at each location there > are 11 independent measurements. No interpolation involved. With the > Bayer sensor there is only one measurement at each location, so in the > 12MP sensor, 3MP of red, 3MP of blue and 6MP of green, and yet in the > final image there are 12MP of red, 12MP of green and 12MP of blue. These > extra values are derived by spatial interpolation and other refinements. that doesn't mean it has 11 times as many pixels. > The SEVIRI more resembles the Foveon approach, but with far better > filtering to separate the individual spectral components. For each > channel, 3 detectors, and 3712 x 3712 pixels. it may be a very accurate pixel, but it's still a *single* pixel and it differs from foveon because each measurement is independent. with foveon, the layers are tightly intertwined. unlike the pretty (and misleading) pictures in their ads, the layers do *not* measure red, green and blue, that's only a result of (here it comes), interpolation. in fact, there is more interpolation with foveon than there is with bayer which is comical, actually. |