From: Martin Brown on
nospam wrote:
> In article <pnczn.155939$y13.49322(a)newsfe12.iad>, Martin Brown
> <|||newspam|||@nezumi.demon.co.uk> wrote:
>
>> There are a handful of cases where the Foveon sensor might give a better
>> image and one of those is when photographing fine black detail on
>> saturated red or blue flowers. Rest of the time it is all marketting.
>
> black detail on saturated colours should be ok with bayer because
> there's a big luminance difference and bayer generally gets that right.

It isn't the Bayer mask that fails in this particular case it is the
chroma subsampling. Try it on a test chart with a Wratten 25 filter and
you will see what I mean. There will be a factor of 2 difference in the
effective resolution between horizontal and vertical in the red channel.

What happens is that the chroma signal for saturated red is allowed to
degrade the luminance channel in all standard JPEG decoders. Some TVs
have started using a more exact method they call 4x4x4 chroma decode.

Foveon save their images as fully chroma sampled JPEGs so the issue of
errors in the subsampled chroma decoding approximations do not arise.

> where bayer has a problem is with two different saturated colours, such
> as red/blue, especially if the luminance is similar. the human eye
> can't handle that particularly well either.

The problem arises later in the imaging chain. Bayer sensor struggles a
bit with a pure red (or pure blue) monochrome images because it has
fewer independent pixels. Normally the luminance channel is able to hide
these defects, but when the situation arises where the luminance channel
is corrupted by the chroma channels then you lose detail.

Regards,
Martin Brown
From: Bubba on
On Apr 20, 3:06 am, Martin Brown <|||newspam...(a)nezumi.demon.co.uk>
wrote:

> I don't think the OP can be helped. He will not help himself by posting
> an image to demonstrate the problem that he thinks he has. It is quite
> possible that the thing he calls red flare is actually a decoding
> artefact in the software that he happens to use!

I corrected myself after having searched the UK forums. It's not "red
flare"; it's "red channel flare."

I haven't posted a picture because there's no need to. I got the
responses I needed by reading these threads. Thanks!
From: Bubba on
On Apr 20, 5:04 am, Martin Brown <|||newspam...(a)nezumi.demon.co.uk>
wrote:
>
> The problem arises later in the imaging chain. Bayer sensor struggles a
> bit with a pure red (or pure blue) monochrome images because it has
> fewer independent pixels. Normally the luminance channel is able to hide
> these defects, but when the situation arises where the luminance channel
> is corrupted by the chroma channels then you lose detail.

Okay, now this interests me. I asked on other threads the obvious
question why a CCD sensor should be considered the equivalent of a
CMOS, if no very low-end P&S has a CMOS.

Why would green not be a problem?
From: Martin Brown on
nospam wrote:
> In article <B6ezn.63223$vC3.49470(a)newsfe04.iad>, Martin Brown
> <|||newspam|||@nezumi.demon.co.uk> wrote:
>
>>>> There are a handful of cases where the Foveon sensor might give a better
>>>> image and one of those is when photographing fine black detail on
>>>> saturated red or blue flowers. Rest of the time it is all marketting.
>>> black detail on saturated colours should be ok with bayer because
>>> there's a big luminance difference and bayer generally gets that right.
>> It isn't the Bayer mask that fails in this particular case it is the
>> chroma subsampling.
>
> bayer samples colour at half the rate of luminance. humans can't see
> the difference, but some think they can (like those who can 'hear'
> differences in speaker cables).

That is a caricature of the situation. It really does depend critically
on what you are trying to image. I actually ran into the problem in a
real world situation. Photographing the transit of mercury with an
H-alpha filter so that prominences were also visible. The planet should
have been a disc but it was obviously distorted to oval.
>
>> Try it on a test chart with a Wratten 25 filter and
>> you will see what I mean. There will be a factor of 2 difference in the
>> effective resolution between horizontal and vertical in the red channel.
>
> if you filter only red, you will reduce the resolution of the sensor
> but that's not a real world scenario.

See above. Astronomers do have filters that are very precisely
monochromatic. Most times we use unfiltered monochrome sensors too.
>
>> Foveon save their images as fully chroma sampled JPEGs so the issue of
>> errors in the subsampled chroma decoding approximations do not arise.
>
> that's a plus but humans can't tell the difference except in extreme
> (i.e., not real-world and contrived) cases.

Red flowers with fine black veins like poppies and tulips is one such
real world case.

>>> where bayer has a problem is with two different saturated colours, such
>>> as red/blue, especially if the luminance is similar. the human eye
>>> can't handle that particularly well either.
>> The problem arises later in the imaging chain. Bayer sensor struggles a
>> bit with a pure red (or pure blue) monochrome images because it has
>> fewer independent pixels.
>
> true but nothing in this world is 'pure saturated red' (or blue or
> green). even bright red objects have a little blue or green in it.

It only has to be close enough that most of the luminance is in the red
channel. Red flowers will cause the problem as do some jazz concerts
with blue and red spotlights.
>
>> Normally the luminance channel is able to hide
>> these defects, but when the situation arises where the luminance channel
>> is corrupted by the chroma channels then you lose detail.
>
> and that only happens in edge cases, like red/blue test charts. that's
> why the foveon fans love those tests, despite it not being relevant to
> real world photography.

Having the sharp edges mangled by subsampling faults stick out like a
sore thumb in the handful of cases where it is relevant.

Claims made for Foveon are completely OTT but there is a small element
of truth in their claims to better colour fidelity. But my point here is
that JPEG decoders are suboptimal in the standard implementation.

Regards,
Martin Brown
From: Alfred Molon on
In article <180420102251244632%nospam(a)nospam.invalid>, nospam says...
> > > > > wrong. every pixel captures luminance.
> > > >
> > > > No. In a Bayer sensor every pixel captures either the red, green or blue
> > > > channel. This is NOT luminance.
> > >
> > > it's one component of the luminance, with the remaining two later
> > > calculated.
> >
> > So you agree that in a Bayer sensor the luminance is not captured in
> > each pixel.
>
> don't twist what i said. luminance is measured at every pixel, even
> though only one component is actually captured.

Luminance requires all *three* colour components. If you do not capture
all three colour components at each pixel, you do not capture luminance
at each pixel.

> > > > > > To properly capture luminance at each pixel you need the full colour
> > > > > > information at each pixel.
> > > > >
> > > > > wrong, as bayer has proven.
> > > >
> > > > No - you are wrong.
> > >
> > > so all of the zillions of photos that very accurately reproduce the
> > > subject all have completely bogus luminance? how can that be?
> >
> > The accuracy is not that high - there are some errors which have the
> > effect of reducing the effective resolution.
>
> it's actually very high, that's why photos look as good as they do.
>
> can you point to a unbiased test (i.e., not from foveon) that shows
> otherwise?

Ever heard of colour aliasing (to make one example)?
--

Alfred Molon
------------------------------
Olympus E-series DSLRs and micro 4/3 forum at
http://tech.groups.yahoo.com/group/MyOlympus/
http://myolympus.org/ photo sharing site
First  |  Prev  |  Next  |  Last
Pages: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
Prev: SI Facescape
Next: FF camera with mirrorless design