OY ... I think I need a Bayer aspirin ...
Nenad Djurdjevic wrote:
>
> Jostein wrote:
>
> > Please someone correct me if I have got this wrong; I'm going to attempt a
> > coherent (don't laugh) description of my own understanding here...
> >
> > IMO, the number of pixels can be considered to be a function of
> > 1. the properties of the CCD used to capture the light, and
> > 2. the software to interprete the captured data.
> >
> > I'm not sure I know all the properties involved, but at least there is the
> > number of physical pixels, and their size (density), and the Bayer
> algorithm
> > involved to reconstruct a colour image from the physical pixels.
> >
> > To consider a 35mm equivalent, assume a 36x24mm CCD. How many pixels do we
> have
> > to put into a 36x24mm CCD to match the resolution power of a film?
> >
> > For simplicity, let's assume that this is equal to the size of film grains
> that
> > we can measure in a microscope. When that is known (i haven't got the
> > number...), we can compare that to the pixel size on the chip. I am pretty
> sure
> > that eg. Provia 100F grain size is smaller than current pixel size in eg.
> > *istD.
> >
> > And here's a problem; there is a physical limit to how small pixels can be
> > before they start to behave erratically. The size used in the *istD is
> thought
> > to be quite optimal, if I have understood this right.
> >
> > If the above is correct, the optimal 36x24mm CCD is one with approximately
> the
> > same pixel density as the *istD has, which would amount to ca. 14
> Megapixels.
> >
> > I don't know if 14 Megapixels in 36x24mm is even close to the grain size
> of
> > ProviaF, but let's assume that it was...:-)
> >
> > Those 14 Mpix must be subjectet to Bayer interpolation before we can see a
> true
> > RGB image. The interpolation will tend to smooth out details that are
> smaller
> > than the smallest cluster of pixels used in the interpolation. I'm not
> sure
> > exatly how this affects the final image quality, but I suspect that this
> is the
> > reason why skin tones and blue skies look smoother with digital, while
> grassy
> > fields or tree branches and twigs look smeared out. Especially in
> low-contrast
> > situations.
> >
> > So, even if 14 Mpix was equivalent to film grain number, CCDs would be
> less able
> > to hold detail (I think this is where the Foveon concept excels over Bayer
> > interpolation).
> >
> > This means that you would need more pixels per area than you have film
> grain per
> > area to record the same level of detail. If film grains can be made
> smaller
> > than CCD pixels, film will win hands down. And that's no matter what
> format you
> > compare.
> >
> > Another discussion is whether the lenses exist to give you full value for
> either
> > recording media. If the practical resolving power is less than pixel/grain
> > density, the difference between film and CCDs is suddenly not _that_
> important
> > any more.
> >
> >
> > So Nenad,
> > I'm not at all sure I know any better than you, I'd rather learn than
> teach. I'm
> > just testing my own grip...:-)
> >
> > Cheers,
> > Jostein
> >
> You're too modest Jostein! You've put together a very coherent and logical
> analysis.
>
> Regards
> Nenad
