comment on COLOR noise in new 5dII

Started Dec 10, 2008 | Discussions
montecotto Regular Member • Posts: 225
comment on COLOR noise in new 5dII

Coming from 40D with 10-22, 17-55/2.8 and 100/2.8 I received my 5DII with 24-105 last week.

I like the feel of the camera, resolution improvement, DOF with my 100/2.8 (85/1.2 would be nicer, sure).

But I'm a bit concerned about noise, specifically color noise at base ISO (luminance noise seems low).

This shot

was taken at 82 mm, 1/160, F/11, ISO 100 (no HTP or ALO), RAW processed in DPP with +0.67 exposure adjustment, standard picture style, slight saturation boost (+2), no NR.

For kicks and lack of better subject matter (there were about 20 min sunshine in between rain and snow) I printed this shot at 7 by 10.5 and saw some purple and cyan discoloration in the print - at first I thought my R1800 had a fit (it's sometimes prissy when it comes to near-grays). But no, it turns out it's not the printer - it's purple/cyan color blotches in the image (200% enlarged for exaggeration).

Am I over-sensitized? Hyper-noise aware? Incompetent? Or is this a result of the lighter colors used in the 5DII sensor's CFA?

Comments appreciated - Thanks!

MC

Luciano Morpurgo Contributing Member • Posts: 710
Re: comment on COLOR noise in new 5dII

Except for few cases at iso 3200 (and a minor part of them), I always had no problem in removing chroma noise without any sacrifice in terms of detail. Putting the chroma NR at 25% - either in DPP or ACR - should do the job.
--
http://www.pbase.com/duca_v2

 Luciano Morpurgo's gear list:Luciano Morpurgo's gear list
Canon PowerShot G5 Canon PowerShot G9 Canon PowerShot S95 Canon EOS 5D Canon EOS 40D +9 more
davejp Regular Member • Posts: 343
ya, that doesn't look so good

ya, that is a bit ugly. Even with the high megapixel handicap, at ISO 100 you should be able to go +1.0 exposure in RAW without killing the shadows with noise. These sorts of things happen when you put too many megapixels on a sensor. Does it look the same without the +2 in saturation?

On a positive note, aside from the color noise you point out, that image looks fantastic for being a 200% crop!

-- hide signature --

EOS 5D & 5D Mark II, and some lenses.
http://dave.jp

(unknown member) Contributing Member • Posts: 873
why don't you use DPP's chroma NR?

montecotto wrote:

Coming from 40D with 10-22, 17-55/2.8 and 100/2.8 I received my 5DII
with 24-105 last week.

I like the feel of the camera, resolution improvement, DOF with my
100/2.8 (85/1.2 would be nicer, sure).

But I'm a bit concerned about noise, specifically color noise at base
ISO (luminance noise seems low).

This shot

was taken at 82 mm, 1/160, F/11, ISO 100 (no HTP or ALO), RAW
processed in DPP with +0.67 exposure adjustment, standard picture
style, slight saturation boost (+2), no NR.

For kicks and lack of better subject matter (there were about 20 min
sunshine in between rain and snow) I printed this shot at 7 by 10.5
and saw some purple and cyan discoloration in the print - at first I
thought my R1800 had a fit (it's sometimes prissy when it comes to
near-grays). But no, it turns out it's not the printer - it's
purple/cyan color blotches in the image (200% enlarged for
exaggeration).

Am I over-sensitized? Hyper-noise aware? Incompetent? Or is this a
result of the lighter colors used in the 5DII sensor's CFA?

Comments appreciated - Thanks!

MC

-- hide signature --

Richard, NC
Never comment on something you don't know about

bionet Senior Member • Posts: 1,133
Re: ya, that doesn't look so good

davejp wrote:

These sorts of things happen when you put too
many megapixels on a sensor.

The 20D/350D did not have this problem, and the pixel pitch is nearly identical.
The 5D2's noise is definitely more ugly with these color blotches.

OP montecotto Regular Member • Posts: 225
Re: ya, that doesn't look so good

davejp wrote:

ya, that is a bit ugly. Even with the high megapixel handicap, at ISO
100 you should be able to go +1.0 exposure in RAW without killing the
shadows with noise. These sorts of things happen when you put too
many megapixels on a sensor. Does it look the same without the +2 in
saturation?

On a positive note, aside from the color noise you point out, that
image looks fantastic for being a 200% crop!

The saturation boost makes the noise slightly more apparent - but "landscape" picturestyle with 0 saturation boost looks even worse.

I agree, detail is otherwise great.

MC

OP montecotto Regular Member • Posts: 225
Re: why don't you use DPP's chroma NR?

One certainly could use chroma noise reduction - though DPP's does not work that well for color blotches (it's great of fine-grained color noise from the 40D).

And I thought that +0.67 exposure boost in a well-lit ISO100 image should not require any color noise reduction - not at this pixel pitch (larger than that of a 40D).

MC

citytrader Regular Member • Posts: 280
Could you post the RAW file!?

Hello, could you post the RAW file!?

Best Regards!

OP montecotto Regular Member • Posts: 225
RAW file posted at sendspace.

Here you go:

http://www.sendspace.com/file/d9ax4z

MC

citytrader wrote:

Hello, could you post the RAW file!?

Best Regards!

halc
halc Regular Member • Posts: 240
Re: comment on COLOR noise in new 5dII

I see what you see and that is the reason I'm disappointed with 5DmkII and money's waiting in the bank account, instead of going towards a purchase (AF issue being the second).

Oak and Acorn Veteran Member • Posts: 6,124
First off, don't "Sat boost" your image for comparison

Boosting the Sat isn't going to help the matter, especially if you haven't applied any Chrominance Noise Reduction to the image.

And second, your 40D image files are less then half the size of the 5DII images. Instead of zooming in on your 40D images to 100% to look for similar color noise blotches, try zooming in on your 40D images to 200% and you'll most likely see color blotches very similar to what you're seeing in your 5DII shots at 100% zoomed in, even at lower ISO's.

I just inspected some low ISO 40D RAW shots in DPP, and at 100% I saw little color noise. But, at 200%, it became quite evident, and very similar in appearance and frequency to your 5DII shot here.

Malcolm L Senior Member • Posts: 1,251
Looks like a reflection of the blue sky.

And the saturation looks high.
Try the grey eyedropper on that area.
Malcolm

 Malcolm L's gear list:Malcolm L's gear list
Canon EOS 5DS R
bionet Senior Member • Posts: 1,133
Re: First off, don't

Oak and Acorn wrote:

And second, your 40D image files are less then half the size of the
5DII images. Instead of zooming in on your 40D images to 100% to
look for similar color noise blotches, try zooming in on your 40D
images to 200%

What for? The 40D sensor has a higher sampling rate (finer pixel pitch), so if anything should be blown up in size, it's the 5D2's shot.

As I said, I do have the feeling that my 350D's images (same pixel pitch) looked better on the pixel level.

Oak and Acorn Veteran Member • Posts: 6,124
Re: if anything should be blown up in size

That was my point. The OP is most likely not looking at his 40D images on a similar scale to the 5DII images. Zooming in to 100% on both the 40D and 5DII files is NOT going to provide identical results when it comes to the appearance of Luminance or Chrominance noise. Of course more noise will be visible in the 5D files (at 100% zoom) due to the massive scale of the image file. IT doesn't mean that the same amount of color noise isn't present in the 40D image file as well, it's just not as visibly noticeable, at a mere 100% zoom, on the smaller 40D image file.

bionet wrote:

Oak and Acorn wrote:

And second, your 40D image files are less then half the size of the
5DII images. Instead of zooming in on your 40D images to 100% to
look for similar color noise blotches, try zooming in on your 40D
images to 200%

What for? The 40D sensor has a higher sampling rate (finer pixel
pitch), so if anything should be blown up in size, it's the 5D2's
shot.

As I said, I do have the feeling that my 350D's images (same pixel
pitch) looked better on the pixel level.

OP montecotto Regular Member • Posts: 225
Re: First off, don't "Sat boost" your image for comparison

Oak and Acorn wrote:

Boosting the Sat isn't going to help the matter, especially if you
haven't applied any Chrominance Noise Reduction to the image.

Of course not - but +2 saturation is something I frequently do with 40D files.

And second, your 40D image files are less then half the size of the
5DII images. Instead of zooming in on your 40D images to 100% to
look for similar color noise blotches, try zooming in on your 40D
images to 200% and you'll most likely see color blotches very similar
to what you're seeing in your 5DII shots at 100% zoomed in, even at
lower ISO's.

I did a fair amount of searching in my 40D files and could not find anything similar at 100% or 200%.

I just inspected some low ISO 40D RAW shots in DPP, and at 100% I saw
little color noise. But, at 200%, it became quite evident, and very
similar in appearance and frequency to your 5DII shot here.

Strongly boosted ISO100 files from the 40D do show color noise as well, but it's very finely grained (not "blotchy").

Best,
MC

citytrader Regular Member • Posts: 280
n/t

n/t

semorg Contributing Member • Posts: 963
seen a similar issu

I made a similar observation on a few other 100% crop images from 5Dii and got accused of "measurebating". The noise I see remind me of the shots I've taken with rebel 400D. I really wish 5D II had stayed at 12mp but allowed for much more shadow noise tolerance.

-- hide signature --

must watch: http://www.climatecrisis.net/
My Latest Gallery: From Peru to Chile: http://www.pbase.com/salim/peruchile2005

Andre Affleck Senior Member • Posts: 2,362
Not the windshield...

You're not talking about the color blotches in the windshield, right? That is simply the grassy landscape coming through the blue windshield. There is some color noise on the silver paint of the car, though. The size of the noise is rather large too. The blotches can span over 40 pixels or more.

rrcphoto Veteran Member • Posts: 6,173
pretty normal.

your black point is set at around 900 for a RAW file, when you slide the exposure up, you are attempting to increase the precision the data, which you can't - there's no data there, so during the demosiacing, you start to get false color, especially for certain tonalities. there is some color noise, but very marginal - I wouldn't call the gray area noise, versus limited bit depth posterization. and then you are compounding the problem by boosting saturation.

that's why expose to the right is always preferable - images will also perform and react better to shadow detail if you drag the exposure down versus up, as you move more precise data down closer to the black point.

OP montecotto Regular Member • Posts: 225
No, Not the windshield...

Andre Affleck wrote:

You're not talking about the color blotches in the windshield, right?
That is simply the grassy landscape coming through the blue
windshield. There is some color noise on the silver paint of the car,
though. The size of the noise is rather large too. The blotches can
span over 40 pixels or more.

No, not the windshield - that's just reflection I agree. It's the paint - the gray-browns that have large blotches - other images with similar colors/brightness have the same problem.

Best,

MC

Keyboard shortcuts:
FForum MMy threads