From: acl on
On Mar 9, 6:11 am, John Sheehy <J...(a)no.komm> wrote:
> "acl" <achilleaslazari...(a)yahoo.co.uk> wrote innews:1173408339.946676.128380(a)v33g2000cwv.googlegroups.com:
>
> > Where did you see this horizontal banding?
>
> It was not readily visible in the D200, for the reason I mentioned. The
> ratio of 2D noise to 1D noise determines how well you can see the 1D noise.

Well yes, but i can clearly see vertical bands and cannot see
horizontal bands at all (and not the high-contrast business, it's
patterned noise).
> I extracted the 1D components from the image, from an unilluminated area,
> and the intensity was actually greater than with my Canon 20D, which has a
> reputation for banding at high ISOs. I subsequently did the same for
> several cameras by various manufacturers, and found that they all had
> banding, and all had stronger banding than the 20D, but their 2D random
> noise was so much more so that the banding was masked.

Well ok I'll look again. I will be surprised if there's are horizontal
banding and I've misses it, but it could be there.

> By banding, I mean offsets in the RAW data on a line-by-line basis, visible
> or not. Even when banding is not visible as "banding", it's removal still
> makes the remaining noise look much more natural (except, of course, for
> the inherent pixel grid of both the signal and noise), despite the fact
> that the standard deviation may drop by less than 1%. Any patterned or 1D
> noise, is much more destructive per unit of standard deviation, than 2D
> random noise. Even binning and downsampling fail to reduce 1D noise at the
> rate it reduces 2D noise; same for viewing full-res images from a distance,
> or printed small; the lines do not fade away, whether they are perceived as
> lines or not.

Yes, I've noticed the same thing.


From: Roger N. Clark (change username to rnclark) on
Paul Rubin wrote:
> "Roger N. Clark (change username to rnclark)" <username(a)qwest.net> writes:
>> Here is pushing some other limits with a Canon 1D Mark II:
>>
>> Night and Low Light Photography with Digital Cameras
>> http://www.clarkvision.com/photoinfo/night.and.low.light.photography
>
> Wow, neat. The super-high-ISO examples have very visible horizontal
> banding--what happens if you take that out with a notch filter?

I have not tried that.

But all this talk about banding is a little
mis-informed in my opinion. John Sheehy seems to be saying that
because there is banding obvious at the low end is evidence for
non-photon noise sources. While true, one must look at the
level of the banding. For example, examine Figure 5 on the above web
page. Set 5 in Figure 5 shows banding at a similar level as the signal
in panels A and B (the left most two squares). But look at the table:
the photons per pixel is only 1.2 in panel A and 0.8 in panel B!
The read noise is 3.9 electrons, so the pattern noise is
about 1/4 the read noise. The problem is that our eyes plus
brain are very good at picking out patterns, whether that pattern
is below random noise, or embedded in other patterns.

It would be interesting to try some filtering on the images.

> Is there a feasible way to remove the Bayer filter from a DSLR sensor?

I do not know.

> What about shorter exposures at super ISO's?

Figure 12 on the above web page is a 1/20 second exposure at equivalent
ISO = 320,000. Do you want faster than that? It is simply a matter of photons
per pixel per exposure. I would not think faster exposures with
similar photons/pixel would appear any different. Longer at lower
light levels would not appear any different either until noise from
dark current starts to show. Dark current noise is the square root
of the dark current, and the 1D Mark II under the temperatures used
was around 0.03 electron/second. So a 10 second exposure would
about a 0.5 electron noise to the 3.9 electron read noise. Thermal
noise equals read noise after about 5 minutes.

Roger
From: acl on
On Mar 9, 2:28 pm, "Roger N. Clark (change username to rnclark)"
<usern...(a)qwest.net> wrote:
> Paul Rubin wrote:
> > "Roger N. Clark (change username to rnclark)" <usern...(a)qwest.net> writes:
> >> Here is pushing some other limits with a Canon 1D Mark II:
>
> >> Night and Low Light Photography with Digital Cameras
> >> http://www.clarkvision.com/photoinfo/night.and.low.light.photography
>
> > Wow, neat. The super-high-ISO examples have very visible horizontal
> > banding--what happens if you take that out with a notch filter?
>
> I have not tried that.
>
> But all this talk about banding is a little
> mis-informed in my opinion. John Sheehy seems to be saying that
> because there is banding obvious at the low end is evidence for
> non-photon noise sources. While true, one must look at the
> level of the banding. For example, examine Figure 5 on the above web
> page. Set 5 in Figure 5 shows banding at a similar level as the signal
> in panels A and B (the left most two squares). But look at the table:
> the photons per pixel is only 1.2 in panel A and 0.8 in panel B!
> The read noise is 3.9 electrons, so the pattern noise is
> about 1/4 the read noise. The problem is that our eyes plus
> brain are very good at picking out patterns, whether that pattern
> is below random noise, or embedded in other patterns.
>

Actually I read his posts as saying, not that photon shot noise is
less important than you say in absolute terms, but that he finds
banding more disturbing. It seems to be a perceptual judgement; he
doesn't appear to be claiming anything quantitatively different from
what you say, just that it bothers him.

For what it's worth, I personally also find patterned noise much more
disturbing than random noise (I really don't mind random noise all
that much unless it gets to very high levels; of course it complicates
my postprocessing but that is another story). It also seems to be the
case that this patterned noise is more obvious to some people than to
others: I have prints from pushed high isos which I find have very
disturbing patterned noise (it jumps out at me immediately, and is
perceptually almost as strong as the image), while my wife and a
couple of friends don't notice it until I point it out, and then seem
to be unaware of it unless they consciously decide to "see" it. I
can't avoid seeing it at all. It seems to depend on the person; maybe
this is part of this confusion (or maybe not).

From: Roger N. Clark (change username to rnclark) on
acl wrote:
>
> Actually I read his posts as saying, not that photon shot noise is
> less important than you say in absolute terms, but that he finds
> banding more disturbing. It seems to be a perceptual judgement; he
> doesn't appear to be claiming anything quantitatively different from
> what you say, just that it bothers him.
>
> For what it's worth, I personally also find patterned noise much more
> disturbing than random noise (I really don't mind random noise all
> that much unless it gets to very high levels; of course it complicates
> my postprocessing but that is another story).

I too agree that pattern noise is more obvious that random noise.
Probably by at least a factor of ten. It is our eye+brain's
ability to pick out a pattern in the presence of a lot
of random noise that makes us able to detect many things
in everyday life. It probably developed as a necessary
thing for survival. But then it becomes a problem when we try
and make something artificial and we see the defects in it.
It gives the makers of camera gear quite a challenge.

Roger


From: John Sheehy on
"Roger N. Clark (change username to rnclark)" <username(a)qwest.net> wrote
in news:45F160FC.5020001(a)qwest.net:


> But all this talk about banding is a little
> mis-informed in my opinion. John Sheehy seems to be saying that
> because there is banding obvious at the low end is evidence for
> non-photon noise sources. While true, one must look at the
> level of the banding. For example, examine Figure 5 on the above web
> page. Set 5 in Figure 5 shows banding at a similar level as the
> signal in panels A and B (the left most two squares). But look at the
> table: the photons per pixel is only 1.2 in panel A and 0.8 in panel
> B! The read noise is 3.9 electrons, so the pattern noise is
> about 1/4 the read noise. The problem is that our eyes plus
> brain are very good at picking out patterns, whether that pattern
> is below random noise, or embedded in other patterns.

Yes, that is a problem, and that is exactly why you can't evaluate noise by
standard deviation alone. It doesn't even take human perception to focus
on the banding; binning and downsampling math focus on it too; an
blackframe from a 20D with 10x the total noise as the horizontal banding
component will show *only* banding noise, and no visible 2D noise at all,
if binned down far enough. I think that this fact speaks volumes as to how
useless standard deviations and S/N ratios based on them can be when
comparing different *characteristics* of noises.


> Thermal noise equals read noise after about 5 minutes.

Statistically, perhaps, but standard deviation does not tell the full
story. You can clearly compare the standard deviations of two noise
situations with the same characteristics, which only vary in terms of
amplitude, but noise comes in a variety of characteristics, and the
standard deviations are not necessarily related to the visual strength of
noise when the characteristics are different. Dark current noise is much
more visible than shot noise, with the same standard deviation, because
most of its energy goes into a minority of pixels.


--

<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
John P Sheehy <JPS(a)no.komm>
><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><