A comparison between the RAW and RGB Histograms in DPP

DPP offers a view on 2 histograms : the RAW histogram and the RGB histogram. One question that often comes to the mind is which one should I use ? and in particular is it better to modify the white/black points in the RGB histogram or in the RAW histogram ?
Some people recommend to not use the RAW histogram because it is hard to control its sliders, in particular the white point slider as the variation is very strong for a small movement. However the following test shows that, with a little patience , the result obtained by modifying the black/white points in the RAW histogram produces better results. I think this is not surprising as the RAW histogram has more information than the RGB histogram.

My personal conclusion is that both histograms are useful : the RAW histogram to make the modifications and the RGB histogram to check the result of these modifications (as the final image will be an RGB image). So the rule could be : "Modify the RAW histogram and Check the RGB histogram"

The following pics present :

• first the original test image
  
• then the image processed by modifying the white/black points of the RGB histogram
  
• and at last the image processed by modifying the white/black points of the RAW histogram
  

• B : is the percentage of black points in the final image
  
• W : is the percentage of white points in the final image
  
• BW : is the percentage of white and black points in the final image (the sum of the previous data)
  
• CG : is the position of the center of gravity of the RGB histogram. A normal exposed image should have its center of gravity at the middle of the range [0,255] ie. 127,5.
  

What a difference! Thanks for bringing this up.

I also feel I need to research DPP more, unsophisticated as it seems to be.

Not sure why you are making distinction between Raw and RGB since the file is a raw RGB file. The only difference between hard presets (the sliders) and the graph is granularity of control. Why not use the curves graph in RGB tab?

jetcode wrote in post #11422437
Not sure why you are making distinction between Raw and RGB since the file is a raw RGB file. The only difference between hard presets (the sliders) and the graph is granularity of control. Why not use the curves graph in RGB tab?

because RAW data is coded on 12 or 14 bits, while in RGB it is on 8 bits. So it is less destructive I think to make the modifications on the RAW histogram rather than on the RGB histogram. See how the colors are more near to the original in the RAW version than the RGB version. I suppose there are also more chances to preserve highlights and shadows by modifying the raw histogram than the RGB histogram.

Raw is raw until you export the file where you select 8 or 16 bit. As far as I know it's the same data set but if that's not the case then they should have 2 exclusive tabs: control for 8 bit and control for 16 bit.

I use a combination of all those controls and export at 16 bits. According to your information this means that some adjustments are done at the 8 bit level and others at the full resolution. In color it may not matter much but in terms of luminance gradation it matters a lot.

There is only one histogram in DPP.

I thought that going to 8 bit, in and of itself, is quite destructive...you're dumping a wide range of colors by the wayside.
Are you suggesting that you're going back and forth between 16 to 8 to 16 and not noticing anything happening?

chauncey wrote in post #11422553
I thought that going to 8 bit, in and of itself, is quite destructive...you're dumping a wide range of colors by the wayside.
Are you suggesting that you're going back and forth between 16 to 8 to 16 and not noticing anything happening?

I don't know the specific internals of the controls themselves but I presume that until you export the file and select resolution everything is done on the raw full range resolution. I believe the sliders are convenient calculations and changes in the curve so the user doesn't have to tweak the histogram directly such as curves in the RGB tab. If not then I've been using the wrong information and Canon is clumsy in not making the distinction more obvious. As it stands I've switched to Lightroom which I assume manipulates the entire data set as is.

Chauncy color is computed as RGB channels which are each 8 (or 16 bit for 48 bit color) bit yielding 24 bit color which has 16 million plus variations. There may be some subtle color that is lost in the conversion but you will never see it on a screen since that gamut is even further reduced in spectrum.

There are 2 histograms in DPP and 2 Tabs : RAW and RGB
If you go to the RGB tab, you'll see the RGB histogram and you'll see that the range of values for this histogram is between 0 and 255, which means 8 bits.
May be I'm wrong but I think that the modifications on the RGB histogram are applied after the data are converted to 8 bits.

I was a little confused about the two histograms existing in the Raw tab controlled by sliders and the histogram end points. Yes the RGB tab has it's own histogram and it appears to be 8 bit which clearly affects how the data is redistributed based on the curve. In other words 8 bit remap of 'n' bit raw data. Curves are about remapping in and out values. So imagine using a 8 bit table to affect a 'n' bit file. You don't lose the resolution of the data you lose resolution of the curve itself.

This is a theory. Again I'd have to see the code to know for sure.

Convetional wisdom is that when setting exposure in camera, one wants RGB histograms so that one can see if any single color is driven to the point of clipping. One would assume that is what you want in setting exposure in RAW adjustment, too.
As for 8-bit vs. 14-bit...the range (min to max) is the same. 14-bit simply gives more steps in between the two extremes. Think of counting 0 to 100, but one allows you to include three decimal values to the right of the decimal point, not merely integers.

Well, by checking again the RGB histogram, I discovered that my histogram was in luminance mode when I have done the modifications. So I switched it to the RGB mode and here is the pic I obtain.

To check if the modifications done on the RGB histogram are done before or after the data are converted to 8 bits, I have converted the image to TIFF 8 bits without modifying the RGB histogram in DPP and then I opened the image in photoshop and adjusted the black and white levels at the same values (73,250) as in DPP RGB version.
And here is the result. Both images look the same which let me think that the modifications on DPP RGB histogram are done after data are converted to 8 bits.

RGB Histogram modified in Photoshop after pic has been converted to TIFF 8 bits

It would be handy to find some real documentation. What I do know is that both Raw and RGB controls affect the output simultaneously. I assume that this is done on raw data and the downscale in resolution occurs on the export function. That premise however requires inside information of the image process itself because the way data is processed does matter.

Hi...What if I use the controles in the RGB tab (that you think is in 8-bit) and then convert the RAW file to a 16-bit TIFF ?

MGH wrote in post #11423105
Hi...What if I use the controles in the RGB tab (that you think is in 8-bit) and then convert the RAW file to a 16-bit TIFF ?

Same result.