21 Apr
2004
21 Apr
'04
6:21 a.m.
On Apr 21, 9:50, Jochen Kunz wrote:
960
On Wed, 21 Apr 2004 00:03:36 +0100 (BST)
ard(a)p850ug1.demon.co.uk (Tony Duell) wrote:
> > Moreover, 3.1M pixels in the camera aren't 3.1M pixels in the
final
> > image. It depends how they're used, but
in the camera, you
> > typically need three pixels, one for each of R, G, and B, to get
one
> > RGB pixel in the image. Some techniques use
even more (the Bayer
> > algorithm uses 4).
> Argh!. You mean they fiddle the figures? I'd assumed that a 'pixel'
> was an RGB triad, not a third of one. So you mean you may only get
1
million points
in the image from a 3.1M pixel camera?
Yes. E.g. with Bayer you have four sub-pixel
per color pixel:
R G
G B
So you get 640 x 480 = ca. 0.3 M "true" color pixels with a 1280 x "Mega pixel" sesor. The image processing
firmware of the camera
interpolates this later to 1280 x 960 RGB pixels.
Exactly, and that's what my (nasty cheap) digital camera does. Looking
at the source for parts of gphoto suggests that's common, so I wouldn't
be at all surprised to find it's the norm, even for higher-end cameras.
If you were a camera manufacturer, and your competitors claimed
1.3Mpixels, would you be more honest and claim 0.4? But I don't
actually know.
--
Pete Peter Turnbull
Network Manager
University of York