UniWB - Practical Use?

[tomczak]

I thought I got it, but apparently I don't...

The idea of UniWB is to set WB so that the RAW RGB 'channels' multipliers are all 1. This way the in-camera R,G,and B histograms (not the luminance) should be able to show those photosites that are actually maxed out in RAW (either R,G or B are clipped). Based on such in-camera histogram, the exposure can be made more agressive, but still without overloading the photosites of any colour.

In PWP RAW, setting WB to 'No WB' is the same as 'Camera WB' which tells me that it worked as intended - the WB multipliers are close to 1. The problem are the histograms: only R histogram are the same in-camera and in PWP, in camera G, and B are not even close (in camera G shows severe clipping, while there is none), which makes the whole techique useless.

What's going on mathematically there?

[keithrj]

I have been following the topic of UniWB for a while and Tonczak's comments just above my post is exactly how I understand the whole thing.

Now here is something else to think about: most of us have heard of 'exposing to the right' but fairly recently there has been some discussion that disputes the whole idea of exposing to the right. Now I have not done any experimenting myself but if the latest thinking is correct then there is no need to fully expose to the right which then makes the whole topic of UniWB a moot point. UniWB is all about exposing to the right using your RAW data and not the histogram produced by the camera (which is based on the fully processed JPG thumbnail).

My thinking is that you should treat each image on its own merit and concentrate on getting your best shot which can be processed later in your favourite program. I will continue to use the histogram as a guide and will adjust accordingly if the histogram is too skewed to one edge or there is clipping (or excessive clipping) where it is un-wanted.

If you are taking photos of static images then why not just bracket and merge for HDR as so many people do?

[keithrj]

Tomzcak,

In PWP RAW, setting WB to 'No WB' is the same as 'Camera WB'

I don't think this is correct but I may be wrong. I believe that 'Camera WB' is what the camera would use when generating a JPG in-camera, i.e. the WB set for the thumbnail. Setting WB to 'No WB' I think is what the RAW would look like without any adjustments for WB.

We also need to be aware of other adjustments like standard curves applied to the image to give more contrast and saturation etc. If we really want to be scientific about the whole issue you need to get your histogram from a fully un-processed RAW file - no WB AND no other adjustments.

Does anyone agree/disagree?

[ksinkel]

I am not sure how valid it is to compare the raw dialog rgb histograms with any histograms the camera displays. The raw dialog histograms are intended to help make adjustments in the dialog. Thus they are computed after each adjustment and can be seen to change each time. Camera histograms are probably also measured after some automatic adjustments have been made and are probably not based on the raw image, though where in the process this may be is pure guesswork. So unless camera and raw dialog adjustments just happen to be the same, the histograms would be different.

On the "No WB" setting, note that the white balance of the light illuminating the image is not known for certain -- thus the starting point for setting the Temp and Tint controls must be based on some assumption. (For example if a feature in an image appears red it could be a white object photographed in red light or a red object photographed in white light or any number of other combinations. We cannot distinguish the cases without making some additional assumption.)

You select the assumption you prefer in Raw Settings. The choices are camera white balance, auto white balance, or no white balance. The Camera WB choice assumes the white balance recorded by the camera is correct. The Auto WB guesses on the correct WB by examining highlights in much the same way as the Color Balance transformation. No WB makes no adjustment whatever -- it simply renders the rgb values as they are in the raw file (after demosaicing). If you notice, under No WB, the Temp and Tint controls have no value because without an assumption about lighting there is no basis for setting them.

Please note, I am not advocating using the No WB setting. It is intended for special purposes only. The best practical setting is definitely Camera WB.

Kiril

[Charles2]

K Sinkel wrote:

I am not advocating using the No WB setting. It is intended for special purposes only. The best practical setting is definitely Camera WB.

Sometimes I follow this procedure:

1. Take a reference shot of a gray card at the scene (mine has six shades of gray), then take the shot of interest.

2. Open the reference shot raw file in the PWP Raw Dialong, choose No WB, and create the TIF for PWP main menu.

3. Open the Color Balance transformation; use probes on the gray card panels to set the white balance. (The first time you do this, it is dramatic to see the color cast on the grays go away and to watch the curve build up; generally it is linear.) Save the curve as a .cb file.

4. Do step 2. on the raw file of interest.

5. Open the Color Balance transformation; Use the saved .cb file.

In this context, it would seem that one should use the No WB setting for both files in the Raw dialog rather than the Camera WB setting - on the principle of minimizing data manipulation. No?

[ksinkel]

I think that what you are doing is sound, though it's more work than you probably need to do. Here is something to try instead:

Open your gray card image in the raw dialog. Use the raw color probe on one of the gray patches to balance the colors. (Admittdly you can only use one gray patch and not several as in the Color Balance dialog. However usually one reference point is all you need.) Make any other raw adjustments you feel are needed. Click OK. This will complete the dialog and set the current raw settings as your default for the next image. (This is assuming it does not have its own wfl file and it is from the same camera.)

Now process the rest of the images in the series. The raw dialog will start at the same settings as you used for the previous image, so if the images are truely all similar, you should not need to make further adjustments.

Kiril

[tomczak]

Would 'No WB' setting be equivalent to have RGB channel multipliers set to 1? I compared the 'No WB' with 'Camera WB' when the camera used UniWB settings and the two seem identical.

Point taken about unknown processing and unequivalency of histograms. But how then do I realize a simple idea: I would like to know if my RAW data are really going to be clipped in any channel or not.

I have a highly sophisticated instrument in my hands with all kinds of histograms, displays and settings, and I almost understand the technology behind RAW and it's processing, and I still can't do that?! This is ludicrous... I figured I have to be able to do something to make histograms useful in setting the exposure.

The UniWB promise was to undo the influence that the WB multipliers have on in-camera JPG and thus should produce histograms that could show the actual RAW clipping. Even if demosaicking and some contrast curves are applied before a JPG is produced, that shouldn't clip a lot more pixels compared to what is actually clipped/not clipped in RAW? I can't figure out what I'm missing...

Hi Charles: why is it better to do the WB in Colour Balance, not in RAW dialog? If you process your gray card in PWP RAW and set the white balance on it, these settings are sticky (i.e. the next RAW you process will have the same WB settings). If the actual WB is set in RAW (i.e. before demosaicking), it is less likely to produce colour artifacts, especially on edges, I believe.

[tomczak]

Here is an 'un-opinionated' article on clipping, histograms and exposure to the right. It's all good, but how do I come to terms with in-camera histograms when shooting RAW?

http://www.cambridgeincolour.com/tutori ... niques.htm

[ksinkel]

Though it is a bit off topic, let's not forget bracketing -- it's usable in many situations and insures both against clipped highlights and insufficient shadow detail.

Kiril

[Charles2]

Maciej wrote:

Hi Charles: why is it better to do the WB in Colour Balance, not in RAW dialog? ... If the actual WB is set in RAW (i.e. before demosaicking), it is less likely to ...

If you watch the PWP Raw Dialog load a raw file, you'll see "demosaicing" before "scaling colors."

As Kiril notes, one gray point is often sufficient, supported by my observation that multiple gray points probed from a gray card shot usually lie on a straight line. Usually. My preference for the Color Balance in PWP "proper" if you will, is because:

• I often process a series of shots over several sessions at the computer, sometimes working on other images in between. Saving and using a .cb file is convenient.
• A strict color balance is not always the best choice; think sunsets, for example. Sometimes I want to apply only 50% or 75% of the balance from a gray card, making the judgment by trial and error in the Color Balance dialog. Similarly, I variously use or do not use the PWP auto white and black points in addition to the gray card points.
• Sometimes the curve from the gray card just fails. I can't always take a shot at the scene that captures the light on the subject 30 meters away.

I have yet to see a dramatic comparison between a scene shot by the UniWB procedure and a shot taken with sensible use of camera feedback (simple chimping; histogram; or blinking shadows and highlights, noted whether they are on irrelevant sky or an important part of the image). Haven't looked assiduously, just haven't seen it.