Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

quantisation/noise in Fuji S7000 photo?

0 views
Skip to first unread message

bugbear

unread,
Feb 6, 2006, 9:53:37 AM2/6/06
to
A friend is trying to make a good still life
photo of a tool he's built.

He's bought a light tent, and 6 table/floor
lamps, tripod, Fuji S7000 camera.

The results are pretty good, although
he's struggling to light the reflective
metal at the same time as the dark wood.

But he's getting HORRID posterisation/quantisation
in the colours.

http://i48.photobucket.com/albums/f234/bugbear33/lower.jpg

I've left the EXIF on the images, although
I sub-sampled 2:1 in Gimp to get under the photo
bucket limit.

Here's a cropped part at full res:
http://i48.photobucket.com/albums/f234/bugbear33/crop.jpg

Any advice on the quantisation would be welcomed.

BugBear

ASAAR

unread,
Feb 6, 2006, 10:41:22 AM2/6/06
to
On Mon, 06 Feb 2006 14:53:37 +0000, bugbear wrote:

> The results are pretty good, although
> he's struggling to light the reflective
> metal at the same time as the dark wood.
>
> But he's getting HORRID posterisation/quantisation
> in the colours.
>
> http://i48.photobucket.com/albums/f234/bugbear33/lower.jpg

The posterization isn't limited to the colors. It's also visible
in the metal parts of the plane, although not as obvious. What does
the histogram show? Perhaps in an attempt to minimize problems with
the metal, the shot was underexposed by several stops? Reducing the
available dynamic range seems like it would contribute to an
increase in posterization. If the exposure is increased to the
point of (or beyond) blowing highlights, does that reduce the
posterization? Reducing the ISO below 200 might also help.

bugbear

unread,
Feb 6, 2006, 11:56:44 AM2/6/06
to
ASAAR wrote:
> On Mon, 06 Feb 2006 14:53:37 +0000, bugbear wrote:
>
>
>>The results are pretty good, although
>>he's struggling to light the reflective
>>metal at the same time as the dark wood.
>>
>>But he's getting HORRID posterisation/quantisation
>>in the colours.
>>
>>http://i48.photobucket.com/albums/f234/bugbear33/lower.jpg
>
>
> The posterization isn't limited to the colors. It's also visible
> in the metal parts of the plane, although not as obvious. What does
> the histogram show?

It appears reasonably continuous.

> Perhaps in an attempt to minimize problems with
> the metal, the shot was underexposed by several stops?

Agreed. The only informations I have is from the EXIF.
But the exposure is "auto", so I guiess we should be OK.


> Reducing the
> available dynamic range seems like it would contribute to an
> increase in posterization. If the exposure is increased to the
> point of (or beyond) blowing highlights, does that reduce the
> posterization?

Exposure should be "there or thereabouts".

> Reducing the ISO below 200 might also help.

Yeah, I thought that, but it turns out the options
are 200,400,800, so he's on "best" already.

BugBear

ASAAR

unread,
Feb 6, 2006, 12:42:06 PM2/6/06
to
On Mon, 06 Feb 2006 16:56:44 +0000, bugbear wrote:

>> The posterization isn't limited to the colors. It's also visible
>> in the metal parts of the plane, although not as obvious. What does
>> the histogram show?
>
> It appears reasonably continuous.

I don't know from "reasonably continuous" what the histogram's
shape actually is. If it's fairly flat, if it has peaks, and if so
where they're located, etc.


>> Perhaps in an attempt to minimize problems with
>> the metal, the shot was underexposed by several stops?
>
> Agreed. The only informations I have is from the EXIF.
> But the exposure is "auto", so I guiess we should be OK.

> . . .


>> If the exposure is increased to the point of (or beyond) blowing
>> highlights, does that reduce the posterization?
>
> Exposure should be "there or thereabouts".

Maybe not. The S7000 probably has several exposure modes, and
each can produce different results. Probably the easiest test would
be to take a few more shots using all of the same settings and modes
already used with one exception. Increase the exposure compensation
by 2 full stops and see if the posterization decreases. And compare
the "before & after" histograms to see if they provide any clues.
The dark wood, even much of the metal appears too dark, which is why
I suspect that underexposure is contributing to the problem.

Marvin

unread,
Feb 6, 2006, 1:27:17 PM2/6/06
to

It looks to me as if the lighting is uneven. The cloth beghind the tool
(it looks like a plane) may get light reflected from the meatl on the back side of the tool.

Bill Tuthill

unread,
Feb 6, 2006, 7:42:37 PM2/6/06
to
bugbear <bugbear@trim_papermule.co.uk_trim> wrote:
> A friend is trying to make a good still life
> photo of a tool he's built.
> He's bought a light tent, and 6 table/floor
> lamps, tripod, Fuji S7000 camera.
> The results are pretty good, although
> he's struggling to light the reflective
> metal at the same time as the dark wood.
> But he's getting HORRID posterisation/quantisation
> in the colours.
>
> http://i48.photobucket.com/albums/f234/bugbear33/lower.jpg
>
> I've left the EXIF on the images, although I sub-sampled
> 2:1 in Gimp to get under the photo bucket limit.

The problem could be JPEG re-writing.

Try to save at the original Q level and chroma subsampling,
which according to the JPEG FAQ minimized artifacting.
I'm sure the Q level is higher than the 80 in your samples.
Chroma subsampling is usually 4:1:1 on most digicams.

bugbear

unread,
Feb 7, 2006, 4:00:19 AM2/7/06
to
ASAAR wrote:
> On Mon, 06 Feb 2006 16:56:44 +0000, bugbear wrote:
>
>
>>> The posterization isn't limited to the colors. It's also visible
>>>in the metal parts of the plane, although not as obvious. What does
>>>the histogram show?
>>
>>It appears reasonably continuous.
>
>
> I don't know from "reasonably continuous" what the histogram's
> shape actually is. If it's fairly flat, if it has peaks, and if so
> where they're located, etc.

I didn't think you were asking about the shape.
The obvious symptom of posterisation would
be a histogram that appeared crenelated.

BugBear

ASAAR

unread,
Feb 7, 2006, 4:40:53 AM2/7/06
to
On Tue, 07 Feb 2006 09:00:19 +0000, bugbear wrote:

>> I don't know from "reasonably continuous" what the histogram's
>> shape actually is. If it's fairly flat, if it has peaks, and if so
>> where they're located, etc.
>
> I didn't think you were asking about the shape.
> The obvious symptom of posterisation would
> be a histogram that appeared crenelated.

Uh, ok, if you say so . . .

bugbear

unread,
Feb 7, 2006, 5:35:26 AM2/7/06
to
bugbear wrote:
> But he's getting HORRID posterisation/quantisation
> in the colours.
>
> http://i48.photobucket.com/albums/f234/bugbear33/lower.jpg
>
> I've left the EXIF on the images, although
> I sub-sampled 2:1 in Gimp to get under the photo
> bucket limit.
>
> Here's a cropped part at full res:
> http://i48.photobucket.com/albums/f234/bugbear33/crop.jpg

DAMN.

It appears from reviews and comments around
usenet and the web that the s7000 is prone
to noise, especially at low res, and especially
when using JPEG, which the camera offers no
control over.

sample:
http://www.dcresource.com/reviews/fuji/finepix_s7000-review/

It looks like the best road to lower noise is to shoot
at 6MP, RAW, and handle the further processing
on the PC, which is inconvenient.

BugBear

0 new messages