CFV Raw File Quality Questions

[bradleygibson]

Saad has sent me a CFV-16 raw file for analysis of a problem with mottled colors.

I took a look and found that this issue is not one of chromatic abberation, or a flaw in the back, but is the result of how (almost) all color imaging sensors use a Bayer filter--sampling only one of (typically) three primary colors at each sensor site.

The other two missing colors must be guessed at to reconstruct a realistic rendering of the original scene. This process is referred to as demosaicing or less formally as 'developing' the picture.

The colors observed here are characteristic of a demosaicing algorithm attempting to produce fine, neutrally colored details, using a colored Bayer filter.

These files are enlarged to 400% make viewing of the problem easier. (Sorry about the file format--this is the only lossless format supported by this board that is true color.)

#1 and #3 show Phocus 1.2 and Lightroom 2.4, respectively with no color filtering. You can easily see the problem around areas of neutral color and high detail.

Images #2 and #4 show the same region but with color filtering for Phocus 1.2 and Lightroom 2.4, respectively. The amount of filtering shown is the maximum for Phocus (so perhaps the software could be updated to allow more filtering), but the setting shown for Lightroom is in fact the default, and can be made stronger if desired.

Hopefully that helps clarify questions around the color issues that were being observed in another thread.

Best regards,
-Brad

 

[jotloob]

Thank you Brad

It sounds all very interesting . Therefore , I would very much like to understand , how to check a RAW file for different parameters and what is color filtering .
Sorry for these stupid questions .

Might even be , that I am not the only dummie here .

Jürgen

 

[polypal]

How are you doing dummie?

Eating on both sides again after your contribution to the yacht of your dentist?



Paul

 

[jotloob]

Paul

I knew something like this would come .

Jürgen

 

[bradleygibson]

Haha--it seems you two know each other well! That's great! Sorry to hear, Jurgen, it sounds like you had some dental work done--that's never fun!

Your question--"how to check a raw file for parameters", I don't think I understand what you are asking. Can you clarify?

As for your other question about what is color filtering, it is basically the art of "smoothing" the colors out. Ideally, just removing the errors from having to "guess" the missing colors I spoke about in my original post. In reality, it's not possible to do this perfectly, so there ends up being some image detail lost when using color filtering. It's best to use as little as possible, but enough to remove any artifacts you find not pleasing.

In Phocus, the control is found under the Adjust tab in the Noise Filter section. It is a slider labeled Color.

In Lightroom, the control is found in the Develop module, under the Detail section in the panel labelled Noise Reduction. It is also the slider labeled Color.

Increasing either of these sliders makes the color smoothing stronger, hopefully eliminating the pesky developing artifacts, but without losing too much detail.

Hope that helps!
-Brad

 

[bradleygibson]

Saad has just provided me with the raw file of the same scene from the CFV-39.

Briefly, the story is as one would expect: without color filtering, color artifacts are quite visible upon close inspection. The CFV-39 fares slightly better in this than the CFV-16, since it is a higher resolution sensor; the 'detailed' information is just a bit easier for the CFV-39 to handle.

Once the color filtering is turned on, the images take on the qualities one would expect.

Here are some lossless .pngs for you to view (the .PDF's are too difficult to preview at 100%). Eagle eyed viewers will notice that the color filtered CFV-16 image also contains JPEG compression artifacts--that was my error--but I don't think it affects the findings much. If you don't know what I'm talking about, please disregard! ).

I've used Lightroom here, only because I am more familiar with it.

Side-by-side comparison without color filtering @ 400%

Side-by-side comparison with color filtering @ 400%

Best regards,
-Brad

 

[saad]

Thanks a lot Brad, so the conclusion is that CFV-39 is better only because of the higher mpx.

 

[bradleygibson]

I'll leave "better" for each person to decide, since factors like budget and other personal preferences will come into play.

From these sample images, I'd conclude that the CFV-39 is higher resolution than the CFV-16, which, of course we already knew.

Other things that jump out at me:
* The CFV's higher resolution is quite significant, to my eye when rendering fine detail (that's good)
* no apparent evidence that the smaller sensels on the CFV-39 have worsened noise performance (not a comprehensive test, but still, it's good)
* default color rendering by Phocus appears more neutral with the CFV-39 (arbitrary, but I still think that's good)

And with the bigger screen and retention of the square option, there seems to be little to sacrifice by going with the CFV-39, except money, of course!

Best regards,
-Brad

 

[wbulte]

I think that we can safely add the significantly better crop factor for wide-angle adepts (like yours truly).

Wilko

 

[bradleygibson]

That's true, too--good point!

 

[jotloob]

Haha--it seems you two know each other well! That's great!

Your question--"how to check a raw file for parameters", I don't think I understand what you are asking. Can you clarify?

As for your other question about what is color filtering, it is basically the art of "smoothing" the colors out. Ideally, just removing the errors from having to "guess" the missing colors I spoke about in my original post. In reality, it's not possible to do this perfectly, so there ends up being some image detail lost when using color filtering. It's best to use as little as possible, but enough to remove any artifacts you find not pleasing.

In Phocus, the control is found under the Adjust tab in the Noise Filter section. It is a slider labeled Color.

In Lightroom, the control is found in the Develop module, under the Detail section in the panel labelled Noise Reduction. It is also the slider labeled Color.

Increasing either of these sliders makes the color smoothing stronger, hopefully eliminating the pesky developing artifacts, but without losing too much detail.

Hope that helps!
-Brad

Yes Brad

We know each other and dislike each other as much as we can .
But there can only be one winner at a time .

As for the clarification .
There was a misunderstanding between me and a friend of mine , who is a NIKON freak .
It must be understood as : checking the RAW file with no parameters set or have them at the default value .
This is regarding USM , and NOISE reduction .

I have played around a bit in PHOCUS with the NOISE REDUCTION (color and luminance slider , as well as moiree) , could see the change in the image and now understand what it is all about .
Thank you for your valuable contribution .

Unfortunately , I can not repeat this in LR2.4 any more .
I can currently not import your supplied 3FR files . Might be it is because I have the ACR5.5(beta) installed , as I was asked by ADOBE people because of the maze pattern issue I described in an other thread .

Regards Jürgen

 

[bradleygibson]

Haha--well you are both great sports! Very entertaining.

As for the files importing into Lightroom, you can convert Saad's .fff's I've posted to .dng's in Phocus. I am sure Lightroom will recognize them then.

Hope that helps,
-Brad

 

[bcphoto]

viewing at 400%

Hi all,

I'm new to this forum and I would like to thank everyone for the useful Hasselblad info. I shoot film on Hassey and digital with Canon. My intent is to move on up to the CFVII back or the CF 39 back.

On this one particular thread I'm puzzled by the use of viewing the images for comparison at 400%. If that is a 400% enlargement from a print I can see the point but I think this is a 400% capture from a digital file. Forgive me if I'm wrong.

100% viewing is one pixel of the source image represented by one pixel on the monitor, or the true image. In PS or other software it is possible to view at 200% or higher. In that case, one pixel of image is on two of the monitor, 400% is one real pixel on 4 pixels of the monitor. So above 100%, the image is being interpolated by the computer's graphics card. It is false information. I see this comparison as being like comparing a grapefruit to a watermelon and then applying that to an apple (the original image). It is pointless.

Please correct me if I'm wrong.

Thanks,

BC

 

[bradleygibson]

Images were posted at 400% to make it easier to see a problem that was being discussed.

Several members noted that they couldn't see the problem when the images were posted at 100%.

 

[bcphoto]

400%

I now see why it was posted that way but it brings up the notion that if it can't be seen at 100% by some or most people then it isn't really much of a problem. And interpolating it to 400% is more than just blowing up the image -- it may be emphasizing or adding to the problem you see.

If that could be seen in a big print I'd be surprised.

Thanks,

BC

 

[bradleygibson]

Hi all,
So above 100%, the image is being interpolated by the computer's graphics card. It is false information.
BC

FYI, if you display a whole number of image pixels (say, 1) to a whole number of screen pixels (say, 2 or 4), it's probably worth pointing out that there is, in fact, no false information being created.

It is the intermediate steps (eg. 123%) and those below 100% (where effectively less than one image pixel is being used) where false information and display artifacts are created.

(But, just so that there's no misunderstanding--I agree; there's not much point in blowing up your images to 400% to look for problems if you can't see them in normal output. It was just done here to help clarify and address a specific question.)

-Brad

 

[bcphoto]

Brad

Brad,

Thanks for the reply. I can see 1 pixel going to 4 easily but more difficult to see it going to 2. Then again I guess that brings in the topic as to whether pixels are square or rectangular (as I've seen someplace).

Anyway I never go beyond 100% for retouching. The whole idea of percentage reminded me of having an apple and eating 100% -- then saying that was only one quarter of it.

BC

 

[fotografz]

Percentages viewed on a computer screen can be false.

I'm working on a 30" monitor, and 100% (or "Actual Pixels") is larger than what a 100% print would be when the image ppi is set for optimal printing on an inkjet printer.

IMO, this is what has led to "pixel peeping" which in turn has lead to over-correction of digital images.

I think that initially one has to print an image to size to evaluate how much correction is really needed and then use that as a base of experience ... if you evaluate it at 100% or more on a computer screen, it often leads to over correction of an issue like the one being discussed here ... and degradation of the image unnecessarily.

BTW, I corrected Saad's image in Phocus exactly the way Brad did ... color noise slider at almost 100%. But that was for on-screen purposes, not for a print.

While I understand Brad's Bayer Filter explanation, what baffles me is that this color noise "artifacts" issue is one I cannot recall experiencing to this degree with any of my digital backs ... as such, I have rarely used the color noise filter in Flex or Phocus, most certainly not at 100% ever. That includes two CFVs, two different 22 meg CFHs (one of which shot DNGs in the camera!), two different 31 meg backs, and two different 39 backs from Hasselblad ... as well as a Leaf Aptus 22, an Aptus 75 and an Aptus 75s.

When I get some time, I will go through my MFD Hasselblad files and see if I overlooked this issue somewhere.

However, I do wonder if something else is at play here? Which is why I was wondering about the heat issue.