Clearing up confusion with Brumbaer/Sinar backs

[Panopeeper]

Brumbaer and eXposure do nothing else then converting and interpolating a RAW image, with applying the CCD's reference files or applying a white shading (if wished), like ANY other back manufacturer are doing (if/when converting to DNG).

As I have been exposed anyway as one, who are analyzing images intently, I looked at the image in question once more. Turned out, that my previous analysis was not intent enough , I have not compared the Brumbaer's and eXposure versions, except for that artificial banding.

Now I took a second look and found, that Brumbaer's carried out a relatively strong noise removal. Although this may reflect the wish of the photographer, it is not something, which is part of everyday's DNG conversion. Furthermore, my form opinion is, that the photographers should be aware (or made aware) of the implicite adjustments the software is doing to their images.

 

[Graham Mitchell]

Panopeeper, have you downloaded the Photoshop document I posted? I wish more people had as it is a lot more useful than looking at one ISO sample, imo.

For example, when you switch between viewing the ISO 50 and ISO 400 layers at 100% you can see the change in noise and detail. From what I have seen there is none of the typical noise reduction look to the ISO 400 file, unlike some Phase samples I have seen which a few people have termed 'painterly'.

If you say that there is some noise reduction going on, I can't confirm or deny that, but if there is, the NR is of a very benign form which does not throw away details.

All I can suggest to everyone is download the PS document and see for yourselves

I challenge owners of other backs to post layered ISO files like this too. IMO, it's the best way to see what's going on.

 

[Panopeeper]

Graham,

I noticed noise reduction only on the Brumbaer converted DNG, not with eXposure.

Noise reduction has to be done in many cases, but that should be on discretional basis. The user (photographer) should decide, if, how much and how needs to be done. NR is inherently connected to loss of detail; sometimes there is a trade-off between smoothness and detail. For example the eMotion 75 image you posted on LL is well served with the NR: the affected (dark) areas are out of focus anyway.

Plus, the DNG conversion is followed by raw conversion anyway, and this is the task of the raw converters. I am firmly convinced, that NR can be done better together with the de-mosaicing and WB application than on its own.

Re the PS file you posted: I prefer to compare the raw files, not the processed ones, particulary when the issue is, what the camera can deliver. There are many raw processors and many options for adjustments, but there is only one original. Moreover, the loss of details can not be judged from the converted files any more.

Gabor

 

[thsinar]

Re: Messed up the posts thoroughly

Dear Panopeeper,

Nobody's offended, except perhaps by the word used: "stink". That is why I intervened. If you would have said "problem" or else it would have been different.

Why it has not been said before is simply because neither Graham nor me had noticed the bug. It is only after asked our SW techs that I got the information and explanation.

And in the meantime I've got some more details about this bug: it was related to the integration, during the conversion process, of the white reference. Under a certain circumstance, when the WR has to be "taken"/"read" from the folder dedicated for it and located on the HD instead of read from the internal storage of the eMotion.

This does not happen with the 6.01 version of eXposure.

Thanks anyway for your time and best regards,
Thierry

Well, this explains all, but it has not been said before here. Graham's Perhaps this was an issue with v6.0? did not indicate to me, that the problem has been analyzed already.

Re the "stink": obviously that was correct, but belated. After further analysis probably I would have suggested, that eXposure makes an error there.

Gabor

 

[thsinar]

No, Panoppeper, that isn't true.

Brumbaer DOES NOT make any NR during conversion, believe me.

If you do not (believe) you can contact Stefan Hess and ask him: Stefan is an open person concerning his tool.

The Brumbaer "DNG Converter" has another integrated feature (which can be enabled or not): the "whitefile creation". This feature does apply a custom "white shading" created by the photographer in case of lens/shift or sensor fall-off. A must feature for architecture photographers (among others). This feature has an option to "Denoise" when used. What does this mean? When creating a white shading, the photographer has to take a shot under the same camera settings as for the image this WR has to be applied to (same f-stop, same shift/tilt/swing) by taking this WR through an opal glass. This opal glass does take away about 2 f-stops of light, thus has to be taken with a longer exposure time of 2 f-stops. This can have the consequence to add noise to this WR file. When later the WR is "subtracted" from the image data, to correct the lens or shift fall-off, this is creating noise in the image. The "Denoise" option is therefore used to eliminate the noise from the WR file, not from the image data itself.

It can now happen, by inadvertently using the "wrong" WR during conversion (this can be chosen in the menu) that some noise is deducted from the image, when it should not. Or this can happen also if the wrong "white ref." file is incorporated.

But in no way does the Brumbaer DNG Converter any NR: if all the parameters are used correctly, with the right sensor white reference, with the correct white shading, etc ..., then the DNGs created by Brumbaer and eXposure are absolutely identical concerning the noise in the image.

I hope this gives some light on how this toll works.

Best regards,
Thierry

Now I took a second look and found, that Brumbaer's carried out a relatively strong noise removal. Although this may reflect the wish of the photographer, it is not something, which is part of everyday's DNG conversion. Furthermore, my form opinion is, that the photographers should be aware (or made aware) of the implicite adjustments the software is doing to their images.

 

[thsinar]

I agree with Graham: a wonderful document to check out what happens between the different layers.

Would effectively be nice to see image samples from other makers here the same way as with Graham's PS document.

But again, I wish to insist here: the Brumbaer DNG converter DOES NOT apply ANY NR.

Best regards,
Thierry

Panopeeper, have you downloaded the Photoshop document I posted? I wish more people had as it is a lot more useful than looking at one ISO sample, imo.

For example, when you switch between viewing the ISO 50 and ISO 400 layers at 100% you can see the change in noise and detail. From what I have seen there is none of the typical noise reduction look to the ISO 400 file, unlike some Phase samples I have seen which a few people have termed 'painterly'.

If you say that there is some noise reduction going on, I can't confirm or deny that, but if there is, the NR is of a very benign form which does not throw away details.

All I can suggest to everyone is download the PS document and see for yourselves

I challenge owners of other backs to post layered ISO files like this too. IMO, it's the best way to see what's going on.

 

[thsinar]

I agree with you, what should be the NR.

Please read what I have written: that is the very reason why there is NO noise reduction going on in Brumbaer.

What can also have happened: when the DNG file is opened in ACR, the "color" noise reduction slider is automatically set to a 25 value, when eXposure leaves this slider at 0.

Thierry

Noise reduction has to be done in many cases, but that should be on discretional basis. The user (photographer) should decide, if, how much and how needs to be done. NR is inherently connected to loss of detail; sometimes there is a trade-off between smoothness and detail.

Plus, the DNG conversion is followed by raw conversion anyway, and this is the task of the raw converters. I am firmly convinced, that NR can be done better together with the de-mosaicing and WB application than on its own.

Re the PS file you posted: I prefer to compare the raw files, not the processed ones, particulary when the issue is, what the camera can deliver. There are many raw processors and many options for adjustments, but there is only one original. Moreover, the loss of details can not be judged from the converted files any more.

Gabor

 

[Panopeeper]

Thierry,

I am not particularly bent on discussing the difference between "noise reduction" and de-noising, and anyway I don't know the precise meaning of "white shading". Instead, I show what differences I see between the Brumbaer and eXposure converted files. If you say that this is not noise reduction, then even better, because I am less than impressed with the effect.

First, the fine histograms; only the first 512 leves, but detailed; eXposure first, Brumbaer second. This is the typical effect of the black level correction, which is the first step of noise reduction (though the raw data of many cameras comes already "corrected"):

Now a selection on a smooth area, again eXposure first:

and some fine, virtually hidden details, eXposure first:

 

[Graham Mitchell]

Panopeeper, what point are you trying to make? My clients don't judge images with RAW analysis software.

 

[Panopeeper]

Graham,

I am showing Thierry, what makes me think that Brumbaer applied some noise reduction (which may have been a mistake on your part, according to Thierry).

My clients don't judge my software based on their knowledge of internals of computer operating systems, but if *I* did not have that knowledge, I could not create the software I am creating.

 

[Graham Mitchell]

Well blasting the forum with meaningless screen grabs isn't going to win you any friends.

No-one is familiar with this software, and whether it is reliable, or your workflow.

The screen grabs don't make much sense anyway. Example: The standard deviations for RGB channels (raw) in the 2 screen shots are identical. This would support the notion that there is no difference in noise levels, or am I missing something?

 

[David K]

This RAW Image Analysis stuff reminds me of that age old philosophy question: If a tree falls in the woods and nobody is there to hear it does it really make a noise... The analogy I'm trying to make is, if you can't see it in the image, who cares... and if you can, that's the best way to show it.

 

[thsinar]

I don't get you, Gabor!

As Graham, my question is the same: what point are you trying to make?

For me and for photographers it is the end result which counts, and those end results have proven to be quite praised concerning IQ.

That's all which counts, not counting the pixels, not analyzing what happens with each of the 33.3 M pixels, etc ...

I respect your "analyzer", I do respect your work, but please refrain from implying or suggesting that we do not know what we are doing. As a matter of fact, I am speaking here for Stefan Hess and Rainer Viertlböck, the 2 who have created and written the Brumbaer tool, when I have not the "right" to speak for them, since Sinar is in NO WAY responsible NOR is Sinar supporting this tool, NOR is it a Sinar application.

Thanks for your understanding and best regards,
Thierry

 

[Dale Allyn]

As one who feels that even pixel-peeping often (inappropriately) overshadows the importance of composition, understanding light, etc., to me this type of RAW file analysis is looking a bit over the top. I care most about the print. If a company is having difficulty delivering good results in the finished image then it makes sense to look for the weak link and try to fix it. On the other hand, if large prints look fantastic I don't care how we got there. That is, I don't think that dissecting algorithms of a successful solution is a good use of energy.

Much of this seems rather academic to me, though may be enjoyable to some I suppose.

 

[Rethmeier]

When I decided to purchase the eMotion-75,which is now the eMotion-75LV,I compared the raws etc with the Aptus and the Phase,the eMotion had the least worked at file
of the 3.
Hence I went for the Sinar.

 

[thsinar]

for those wishing the prove that what I said below is true (no NR in the DNG conversion), please read under the following link:

http://luminous-landscape.com/forum/index.php?showtopic=25473&st=60

post N° 61

I think it makes it clear, after Stefan Hess' explanation about his Brumbaer tool: it is exactly what I have written and explained below, there is no NR in the Brumbaer DNG conversion, never (nor in Sinar eXposure).

Best regards,
Thierry

 

[PeterA]

This type of analysis is interesting to me for a number of reasons - the least of which has any relevancy to photography especially where it matters - the finished print...

however some people have an analytical and empirical 'bent'..the technology in a MFD back ( or any digi camera ) is fascinating in itself - yes?

nothing wrong with an inquiring mind chaps ! Still I too would like to know what the point being made is ..and then how it links into a bigger picture that may be of relevance to more people in their decision making or approach to workflow..

 

[Guy Mancuso]

I have to agree with Peter and ask the question when does all this analyzing make sense to the shooter. At what point does it become relevant that i shoot a certain way or at a certain setting to squeeze every drop from the back and it's software. I understand the limits and such of these backs and I guess this begs the question how far can we stretch something until we actually fall off the cliff. Just like using USM when processing we need to know at what limit there is so it still looks good before over sharpening a file. So i understand the science of it or the analysis but at some point I need something whispered in my ear , hey Guy don't do that or you will get in trouble.

 

[brumbaer]

As I already said on the LL forum.

eMotionDNG does not do any dedicated sharpening or noise reduction.

The differences in "brightness" between Sinar Software and the Brumbaer Tools are created by different approaches how to "scale" the pixel values.

My approach is different than the XPosure/CaptureShop-approach. The reason is that my software was developed completely independent and without help from Sinar. I stress this point, because it is the single most important factor for: "why the approaches are different".
When you get a NEF file or a CRW half of the processing that the Tools do is already done.

Instead of proclaiming a "stink all over the world", you should have considered the case that I do something right and the other guys do something less right (does this make it left ?). Or at least that I do something different instead of cheating.


Regards
SH

 

[thsinar]

Thanks Stefan!

I was not expecting less as an answer and it puts things definitively in the right perspective.

Best regards,
Thierry

As I already said on the LL forum.

eMotionDNG does not do any dedicated sharpening or noise reduction.

The differences in "brightness" between Sinar Software and the Brumbaer Tools are created by different approaches how to "scale" the pixel values.

My approach is different than the XPosure/CaptureShop-approach. The reason is that my software was developed completely independent and without help from Sinar. I stress this point, because it is the single most important factor for: "why the approaches are different".
When you get a NEF file or a CRW half of the processing that the Tools do is already done.

Instead of proclaiming a "stink all over the world", you should have considered the case that I do something right and the other guys do something less right (does this make it left ?). Or at least that I do something different instead of cheating.


Regards
SH