A1-Noise Filtering Technique: Neat Image & Noise Ninja

On the Mac platform you should try following ones ...

Grain Surgery (www.visinf.com)
dFine (www.nikmultimedia.com) (not yet for A1 .... but also for windows)
Power Retouche (www.powerretouche.com)
Intellihance (www.extensis.com, also windows)

I use myself dFine for a while for Nikon 990. Works pretty well. Grain surgery is also a great filter.
Could anyone comment on the efficacy of the noise reduction
features of Photoshop CS (when editing RAW images) versus other
programs and plugins?

I have only used CS with my new A1. I really, really like shooting
and editing in RAW. But are there other options that are markedly
better for noise reduction?
 
using the technique that I outline here:

http://www.objective.no/geirove/web/new_page_1.htm

down on the page under the text:
"For the shots that still have visible Noise"

youl will find a short explanation.

I have used it a lot: It works remarkably well.

Geir Ove
Could anyone comment on the efficacy of the noise reduction
features of Photoshop CS (when editing RAW images) versus other
programs and plugins?

I have only used CS with my new A1. I really, really like shooting
and editing in RAW. But are there other options that are markedly
better for noise reduction?
 
No worries.

Maintaining detail is a high priority. If you remove too much colour noise then the remaining luminosity noise looks "strange". When producing prints colour noise is much harder to see than on-screen.

Finally, it's best to balance noise profiles for daylight outdoor pictures.

Jawed
 
Hello,

I have not so far seen any negative effects of using high Cr and Cb filtering. I have tried it on Natural Light pictures as well. Will do some more testing though! I see that most of the "Retain Natural Look" filters that come with Neat Image aslo does this.

Excerpt from the Neat Image docs:

"As human vision is not very sensitive to variations of colors, strong filtration in the Cr and Cb channels does not noticeably distort an image, but efficiently removes color noise. "

And that has been my experience so far.

Geir Ove
No worries.

Maintaining detail is a high priority. If you remove too much
colour noise then the remaining luminosity noise looks "strange".
When producing prints colour noise is much harder to see than
on-screen.

Finally, it's best to balance noise profiles for daylight outdoor
pictures.

Jawed
 
You are missing my point. By strongly filtering colour noise, you leave behind artefacts of noise in the luminosity channel (if you want to retain any detail). These artefacts look WORSE than if you had weakly filtered colour noise. The colour noise "hides" the luminosity noise. The whole point is to produce a balanced-looking image, rather than just flattening the fúck out of all noise possible.

The default settings of NI are far too strong. Why the developer continues to peddle his software so BADLY defaulted mystifies me, as while it might produce impressive-looking results for blondes, to the average careful photographer who's interested in noise reduction it is much too heavy handed. The comments on the eye's reduced sensitivity to Cr and Cb noise are all very well, but that doesn't give licence to destroy all. I'm sure this is why many people try NI and reject it straight away, because it produces the infamous "plastic look".

When you've tested (published results from) your filter settings on more than a single, artifiical, sample picture then I'll have more interest in your settings. I have spent a lot of time testing my settings on numerous images. I've been fiddling with NI off and on for around two years.

Put bluntly, your filter settings are, to my eyes, no more useful than the default configuration of NI. That said, I applaud the fact that you are bringing people's attention to the filter settings, and not just to the noise profiles.

For people who can't be bothered with all this, there's always Noise Ninja. Hahaha. NN is of no interest to me until someone comes out and says it produces dramatically better results. NI sets the standard when used properly. I'm a geek, I know how to use NI properly.

The reason I'm not publishing the results of my tests is simple: My A1 produces horrendous noise, making it completely un-typical. Therefore all my testing has been with sample pictures from DPReview, Imaging Resource and Steve's Digicams. Due to copyright restrictions I will not post noise-filtered versions of these sites' pictures. When I get an A1 that works properly, then maybe I'll publish pix.

Jawed
I have not so far seen any negative effects of using high Cr and Cb
filtering. I have tried it on Natural Light pictures as well. Will
do some more testing though! I see that most of the "Retain
Natural Look" filters that come with Neat Image aslo does this.

Excerpt from the Neat Image docs:

"As human vision is not very sensitive to variations of colors,
strong filtration in the Cr and Cb channels does not noticeably
distort an image, but efficiently removes color noise. "

And that has been my experience so far.

Geir Ove
No worries.

Maintaining detail is a high priority. If you remove too much
colour noise then the remaining luminosity noise looks "strange".
When producing prints colour noise is much harder to see than
on-screen.

Finally, it's best to balance noise profiles for daylight outdoor
pictures.

Jawed
 
Hello,

Jawed said: "The whole point is to produce a balanced-looking image, rather than just flattening the fúck out of all noise possible."

" Put bluntly, your filter settings are, to my eyes, no more useful than the default configuration of NI. "

That's simply not true; your strong and undocumented argumentation is not a goo basis for a good discussion.

No reason start using the F* word all over the place.

When I have the time, I will try to publish more examples. I have used hours to test my settings so far, and have NOT found that they produce any adverse effects and are easy on the details.

I tried to get some intelligent discussion on the subject, but it seems like you take this personally and start yelling as soon as I present some counter-arguments. Not the kind of discussion I wanted.

I AM looking to impove my results. I have test printed pictures on a Canon 9000 printer and the results are very good so far.

Geir Ove
The default settings of NI are far too strong. Why the developer
continues to peddle his software so BADLY defaulted mystifies me,
as while it might produce impressive-looking results for blondes,
to the average careful photographer who's interested in noise
reduction it is much too heavy handed. The comments on the eye's
reduced sensitivity to Cr and Cb noise are all very well, but that
doesn't give licence to destroy all. I'm sure this is why many
people try NI and reject it straight away, because it produces the
infamous "plastic look".

When you've tested (published results from) your filter settings on
more than a single, artifiical, sample picture then I'll have more
interest in your settings. I have spent a lot of time testing my
settings on numerous images. I've been fiddling with NI off and on
for around two years.

Put bluntly, your filter settings are, to my eyes, no more useful
than the default configuration of NI. That said, I applaud the
fact that you are bringing people's attention to the filter
settings, and not just to the noise profiles.

For people who can't be bothered with all this, there's always
Noise Ninja. Hahaha. NN is of no interest to me until someone
comes out and says it produces dramatically better results. NI
sets the standard when used properly. I'm a geek, I know how to
use NI properly.

The reason I'm not publishing the results of my tests is simple: My
A1 produces horrendous noise, making it completely un-typical.
Therefore all my testing has been with sample pictures from
DPReview, Imaging Resource and Steve's Digicams. Due to copyright
restrictions I will not post noise-filtered versions of these
sites' pictures. When I get an A1 that works properly, then maybe
I'll publish pix.

Jawed
I have not so far seen any negative effects of using high Cr and Cb
filtering. I have tried it on Natural Light pictures as well. Will
do some more testing though! I see that most of the "Retain
Natural Look" filters that come with Neat Image aslo does this.

Excerpt from the Neat Image docs:

"As human vision is not very sensitive to variations of colors,
strong filtration in the Cr and Cb channels does not noticeably
distort an image, but efficiently removes color noise. "

And that has been my experience so far.

Geir Ove
No worries.

Maintaining detail is a high priority. If you remove too much
colour noise then the remaining luminosity noise looks "strange".
When producing prints colour noise is much harder to see than
on-screen.

Finally, it's best to balance noise profiles for daylight outdoor
pictures.

Jawed
 
LOL you think I was annoyed?
Hello,

Jawed said: "The whole point is to produce a balanced-looking
image, rather than just flattening the fúck out of all noise
possible."

" Put bluntly, your filter settings are, to my eyes, no more useful
than the default configuration of NI. "

That's simply not true; your strong and undocumented argumentation
is not a goo basis for a good discussion.
I looked at your pictures and I came to an opinion. My opinion is CORRECT since it is nothing more than an OPINION. You said yourself that your pictures have less detail with your NI configuration than with my NI configuration. You also said that your pictures have less colour noise with your NI configuration than with mine. My response is that, IN MY OPINION, your NI configuration is too strong.
No reason start using the F* word all over the place.
Hahaha, bit strong for you, lol.
When I have the time, I will try to publish more examples. I have
used hours to test my settings so far, and have NOT found that they
produce any adverse effects and are easy on the details.
That's fine. Try this set of pictures and see if you think your filter settings are appropriate:

http://www.imaging-resource.com/PRODS/A1/A1FAR100.HTM
http://www.imaging-resource.com/PRODS/A1/A1FAR200.HTM
http://www.imaging-resource.com/PRODS/A1/A1FAR400.HTM
http://www.imaging-resource.com/PRODS/A1/A1FAR800.HTM

Click on the thumbnail to download the full picture.
I tried to get some intelligent discussion on the subject, but it
seems like you take this personally and start yelling as soon as I
present some counter-arguments.
No you didn't present a counter argument. You simply stated your preference. I re-stated my preference and why it is that I find your configuration unacceptable. I find your approach too similar to the default behaviour of NI, which I particularly dislike.
Not the kind of discussion I wanted.
Ah well, never mind. If you can't take the heat. Though nothing I said was insulting - maybe English isn't your native tongue.
I AM looking to impove my results. I have test printed pictures on
a Canon 9000 printer and the results are very good so far.
The thing I've learnt about NI is that it is very easy to be suckered into thinking you've got a good configuration. Over the two years I've been playing with it, I've learnt that it takes a lot of testing to come up with a robust profile and filter. Sure, the chances are that you will tweak the settings by hand for each picture - I'm a geek, frobnicating geeky things like NI is an interesting diversion once in a whle...

I'm hoping that by providing the filters it will give people a clue as to how to fine-tune NI in the most effective way. As I said earlier, I applaud your action in bringing people's attention to the Filter part of NI, a part of NI that people barely recognise is there.

Jawed
 
Hello,

All I wanted was a contructive discussion, not ....

My ISO 100 Preset does not remove detail that I can see, but indeed removes color noise when present.

My ISO 200 - 800 has a bit too strong detail filtering, but is far** nicer on the details than the Neat Image "Default".

As a result, I have reduced the strength of filtering for High and Medium Freq. for ISO 200 to 800, and they now look pretty much like yours in the detail department, but still remove more color noise (when present as on my TV screen shot)

It is interesting to dicuss how to improve these settings, as the results of it is quite amazing :-)

Geir Ove
Hello,

Jawed said: "The whole point is to produce a balanced-looking
image, rather than just flattening the fúck out of all noise
possible."

" Put bluntly, your filter settings are, to my eyes, no more useful
than the default configuration of NI. "

That's simply not true; your strong and undocumented argumentation
is not a goo basis for a good discussion.
I looked at your pictures and I came to an opinion. My opinion is
CORRECT since it is nothing more than an OPINION. You said
yourself that your pictures have less detail with your NI
configuration than with my NI configuration. You also said that
your pictures have less colour noise with your NI configuration
than with mine. My response is that, IN MY OPINION, your NI
configuration is too strong.
No reason start using the F* word all over the place.
Hahaha, bit strong for you, lol.
When I have the time, I will try to publish more examples. I have
used hours to test my settings so far, and have NOT found that they
produce any adverse effects and are easy on the details.
That's fine. Try this set of pictures and see if you think your
filter settings are appropriate:

http://www.imaging-resource.com/PRODS/A1/A1FAR100.HTM
http://www.imaging-resource.com/PRODS/A1/A1FAR200.HTM
http://www.imaging-resource.com/PRODS/A1/A1FAR400.HTM
http://www.imaging-resource.com/PRODS/A1/A1FAR800.HTM

Click on the thumbnail to download the full picture.
I tried to get some intelligent discussion on the subject, but it
seems like you take this personally and start yelling as soon as I
present some counter-arguments.
No you didn't present a counter argument. You simply stated your
preference. I re-stated my preference and why it is that I find
your configuration unacceptable. I find your approach too similar
to the default behaviour of NI, which I particularly dislike.
Not the kind of discussion I wanted.
Ah well, never mind. If you can't take the heat. Though nothing I
said was insulting - maybe English isn't your native tongue.
I AM looking to impove my results. I have test printed pictures on
a Canon 9000 printer and the results are very good so far.
The thing I've learnt about NI is that it is very easy to be
suckered into thinking you've got a good configuration. Over the
two years I've been playing with it, I've learnt that it takes a
lot of testing to come up with a robust profile and filter. Sure,
the chances are that you will tweak the settings by hand for each
picture - I'm a geek, frobnicating geeky things like NI is an
interesting diversion once in a whle...

I'm hoping that by providing the filters it will give people a clue
as to how to fine-tune NI in the most effective way. As I said
earlier, I applaud your action in bringing people's attention to
the Filter part of NI, a part of NI that people barely recognise is
there.

Jawed
 
The updated versions are leaner on details from ISO 200 and up.
The updated versions also have a version with sharpening.

download here:

http://www.objective.no/geirove/web/noise_reduction.htm

and give me feedback :-)

Geir Ove
Hello,

I have just updated my Photo Web :-)
It's here: http://www.objective.no/geirove/web .

I have a section where you can download optimum ( ) Neat Image
"Filter Presets" for each ISO Profile! For optimum use of Neat
Image, you must not only match the correct ISO Profile to the
Image. You must also tune the "Noise Filter Settings".
Neat Image provides some nice presets. Hower, I believe I have
found better settings: You can see the effect of these settings on
this page:

http://www.objective.no/geirove/web/noise_reduction.htm

and also download my Filter Presets. I have used many days of
testing to finally end up with these settings .. puh !

I believe the results are very** good: Look for yourself on my
web pages.

My Filter Presets currently have no Sharpening: I will probably
provide Filer Presets with sharpening later.

Noise Ninja:

I haven't had the time to work much with Noise Ninja yet, but I
belive the program has potential to produce just as good results: I
will test this later, but it is very time consuming !

**
If someone finds BETTER Filter Presets for each A1 ISO Profile,
please make them available !

Geir Ove
 
Hope you're still here Jawed.

Jawed wrote:
(snip)
The reason I'm not publishing the results of my tests is simple: My
A1 produces horrendous noise, making it completely un-typical.
So we're back in the noise business, and I'm even participating (shudder!) :-)

But Jawed, I have a question that I haven't actually found answered anywhere. Maybe it's there, but I don't recall having seen it. I'm not asking Geir Ove, since he owns The Perfect Camera :-)))

Would you be bothered to shoot some noisy comparison samples where one is 5Mpix and the other 3Mpix out of the camera , and post some 100% crops?

Yes, I know perfectly well that the camera is a 5Mpix camera, but I might be willing to accept a camera that takes good pictures at 5Mpix in good light and good pics at 3Mpix in less than perfect light conditions. Your results might be transferrable to the Sony F828. After all, a 2/3" chip at 5Mpix is trying to do the impossible.

Vegard
 
Hi,

Thanks for posting these, I'll be sure to download them and have a look a little later when I have a spare moment.

I was just wondering, are these presets constructed from JPEG images? If so, at what quality level? Or RAW files? Converted using which converter and interpolation method?

I tend to use RAW and Dalibor's plugin, and usually stick to the default Colour Correction interpolation method.

I find that adjusting exposure compensation in the RAW converter changes the distribution of noise levels in the image, so presets are of limited use. It also stops NN working. I'm assuming that using a different RAW converter has a similar (but less drastic) effect.

Cheers,
Jon
--
A1 in Cuba gallery 2003: http://photos.jonread.com/Cuba/
Hot Pixel noise remover: http://www.jonread.com/hotpixel/
 
The A1's re-sampling algorithm, that sizes down from 5MP to 3MP is poor quality, in my opinion.

In my tests with my D7, resizing from 5MP to 2MP (the D7 didn't have a 3MP picture size) produces MORE noise, not less.

It is better to shoot at the highest resolution and use post-processing if you want to achieve the best noise levels under difficult conditions.

Under difficult conditions your enemy is JPEG compression. JPEG compression converts noise into something else, best described as "mush". By using the highest-quality JPEG compression (or indeed shooting TIFF or RAW) you will produce a picture with noise that cleans up better in post-processing.

On the D7, TIFF and RAW modes produce less noise in pictures than JPEG Fine (which is the best quality JPEG mode on the D7. RAW is the best overall for the D7. On the A1 it seems (from Phil's test) that RAW produces more noise than JPEG. Both modes produce different kinds of artefacts, neither of which are desirable. You pays your money and takes your choice as they say. My life is too short (I take too many pix) to use RAW.

It is interesting to me that the conversion of RAW files used by Dimage Viewer appears to use the OLD algorithms from the D7 series, when converting A1 pictures. A1 RAW pictures look noticeably different from A1 JPEG pictures, and not entirely better, in my view.

So, in summary, if you want to maximise picture quality, and the ease with which you can post-process for noise, then shoot at maximum resolution with maximum quality. In the A1, RAW is not definitively the maximum quality, which is a shame really. JPEG Extra Fine is very good - it's very hard to find any difference between Extra Fine and Super Fine (TIFF).

When I have an A1 that works properly, perhaps I can give specific answers based on my A1. Until then, my experience with the D7 will have to do, lol. But I think the A1's 3MP mode is bad - avoid it. I've done a lot of testing with this mode and I won't be using it out of choice. The problems that I have with it are not to do with noise, but are to do with a significant increase in colour moire artefacts. These artefacts become so strong that they are clearly visible in 25% view, which is depressing really.

Minolta's image processing techniques continue to lag competitors in their sophistication, in some important ways. The other side of the coin is that I still prefer A1 pictures over ALL competing 5MP cameras, bar NONE - so their priorities obviously lie in different places to competitors.

Jawed
Jawed wrote:
(snip)
The reason I'm not publishing the results of my tests is simple: My
A1 produces horrendous noise, making it completely un-typical.
So we're back in the noise business, and I'm even participating
(shudder!) :-)

But Jawed, I have a question that I haven't actually found answered
anywhere. Maybe it's there, but I don't recall having seen it. I'm
not asking Geir Ove, since he owns The Perfect Camera :-)))

Would you be bothered to shoot some noisy comparison samples where
one is 5Mpix and the other 3Mpix out of the camera , and post some
100% crops?

Yes, I know perfectly well that the camera is a 5Mpix camera, but I
might be willing to accept a camera that takes good pictures at
5Mpix in good light and good pics at 3Mpix in less than perfect
light conditions. Your results might be transferrable to the Sony
F828. After all, a 2/3" chip at 5Mpix is trying to do the
impossible.

Vegard
 
Hello,

The Filter Presets has nothing to do with JPEG etc. The ISO Profiles may vary according to JPEG compression etc. The ISO Profiles for the A1 was (on the Neat Image Site) was developed by me using JPEG Fine level.

However, my Filter Presets try not to filter out every bit of noise, but be ""nice" on the details in the picture. By doing this, the whole setup gets less "critical": The little variation that exists between images will then mean that some image will have a little (very little) more noise and some a little less. But so what? The npise levels left after filtering ISO 100 and 200 pictures are very low anyway.

If Filter Presets are tuned to the Max for a certain type of pictures, they will easily be far to strong for other pictures. I have done my best to avoid that my Filter presets behaves like that.

Geir Ove
Hi,

Thanks for posting these, I'll be sure to download them and have a
look a little later when I have a spare moment.

I was just wondering, are these presets constructed from JPEG
images? If so, at what quality level? Or RAW files? Converted
using which converter and interpolation method?

I tend to use RAW and Dalibor's plugin, and usually stick to the
default Colour Correction interpolation method.

I find that adjusting exposure compensation in the RAW converter
changes the distribution of noise levels in the image, so presets
are of limited use. It also stops NN working. I'm assuming that
using a different RAW converter has a similar (but less drastic)
effect.

Cheers,
Jon
--
A1 in Cuba gallery 2003: http://photos.jonread.com/Cuba/
Hot Pixel noise remover: http://www.jonread.com/hotpixel/
 
Long story, but I also had the opportunity to compare the two: One generated more noise than the other at LONG exposures (10 - 30 sec). Both produces very good results at 1/10 sec and shorter shutter times, but the best one was also somewhat better in that department.

Yes, the one I have now is VERY good: Even ISO 200 looks very good!

Jawed: If your camera is horrendous: Why don't you ask Minolta to replace it? It should be no problem AT ALL!

Geir Ove
Jawed wrote:
(snip)
The reason I'm not publishing the results of my tests is simple: My
A1 produces horrendous noise, making it completely un-typical.
So we're back in the noise business, and I'm even participating
(shudder!) :-)

But Jawed, I have a question that I haven't actually found answered
anywhere. Maybe it's there, but I don't recall having seen it. I'm
not asking Geir Ove, since he owns The Perfect Camera :-)))

Would you be bothered to shoot some noisy comparison samples where
one is 5Mpix and the other 3Mpix out of the camera , and post some
100% crops?

Yes, I know perfectly well that the camera is a 5Mpix camera, but I
might be willing to accept a camera that takes good pictures at
5Mpix in good light and good pics at 3Mpix in less than perfect
light conditions. Your results might be transferrable to the Sony
F828. After all, a 2/3" chip at 5Mpix is trying to do the
impossible.

Vegard
 
I've waited a while so that if there's bad stock here in the UK, I'm minimising my chances of getting a bad one in replacement - if they've improved quality control, that is (sigh). That's if they replace rather than repair. Also, I've needed the camera. I shoulda done it this week but I've been caught up with other things. Next week, maybe.

Jawed
Long story, but I also had the opportunity to compare the two: One
generated more noise than the other at LONG exposures (10 - 30
sec). Both produces very good results at 1/10 sec and shorter
shutter times, but the best one was also somewhat better in that
department.

Yes, the one I have now is VERY good: Even ISO 200 looks very good!

Jawed: If your camera is horrendous: Why don't you ask Minolta to
replace it? It should be no problem AT ALL!

Geir Ove
 
Hello,

Strange thing is: The second (very good ) camera that I got, has an earlier serial # than the first one I got ! So there seems to be some variation at least for the first samples released on the market. Maybe, as you say, Minolta has improved QC by now.

I do hope you get a GOOD A1; They are, IMO, simply incredible !

Geir Ove
Jawed
Long story, but I also had the opportunity to compare the two: One
generated more noise than the other at LONG exposures (10 - 30
sec). Both produces very good results at 1/10 sec and shorter
shutter times, but the best one was also somewhat better in that
department.

Yes, the one I have now is VERY good: Even ISO 200 looks very good!

Jawed: If your camera is horrendous: Why don't you ask Minolta to
replace it? It should be no problem AT ALL!

Geir Ove
 

Keyboard shortcuts

Back
Top