Pages: [1]   Go Down

Author Topic: Problem adjusting white point in post-processing  (Read 5215 times)

Hening Bettermann

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 945
    • landshape.net
Problem adjusting white point in post-processing
« on: December 27, 2010, 03:17:13 pm »

Hi!

hope you had a nice Christmas.

I have 2 questions concerning ETTR and post-processing.

Here are 2 raw images, exposed almost 2 f-stops apart (1.6 and 6 seconds), displayed linear.

Originally, I intended to stack these. But then I discovered that BOTH needed the SAME adjustment of the white point in Raw Developer to become "clip-free" in the highlights, namely 255/254. So the light image would contain all the same information with the bonus of less shadow noise. So why bother stacking?
(I have recently learned that the benefit of ETTR above base ISO is a misunderstanding, but this is not the issue here).

But no.

The overall lightness of the dark image conforms to my memory, as noted when that memory was fresh. (With a very slight  S-curve, it is close to my memory and "rendering intent".)

Now, shouldn't I be able to make the linear light image look like the linear dark image just by adjusting the white point, but staying linear? But I can't, neither in Raw Developer nor on the linear TIF in my image editor, PhotoLine.
As soon as I lower the y-value of the white point, the sky turns muddy, whereas it is clear in the dark image.

Furthermore, I can not by ANY form of the TRC make the light image look like the dark one in overall lightness. When I lower the mid tones to a degree in the nearness of the intended overall lightness, the image becomes over-contrasty and the colors over-saturated. No wonder, looking at the TRC that is required to achieve the lowering of mid-tones.

So the questions are:

1- How can it be that 2 images exposed 2 EV's apart can require the SAME adjustment of white point to just avoid highlight clipping?

2- Why can I not adjust lightness of the 2 images, staying linear?

Any enlightenment in the dark period of the year?

Thank you for your interest.

« Last Edit: January 07, 2011, 08:27:56 pm by Hening Bettermann »
Logged

nilo

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 397
Re: Problem adjusting white point in post-processing
« Reply #1 on: January 05, 2011, 03:30:30 pm »

I can't help you, but I would like to know if there is any news on that matter?

Don't forget to keep us posted (and maybe somebody still stumbles across your post after this "dark period of the year").

regards
Logged

Hening Bettermann

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 945
    • landshape.net
Re: Problem adjusting white point in post-processing
« Reply #2 on: January 05, 2011, 04:06:52 pm »

Hi Nino,

thank you for your concern. The news is that I ended up stacking these 2 images and did all right, but still, I would like to understand why I could not use the light version alone.

Good light! - Hening.

Peter_DL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 544
Re: Problem adjusting white point in post-processing
« Reply #3 on: January 05, 2011, 05:37:30 pm »

... shouldn't I be able to make the linear light image look like the linear dark image just by adjusting the white point, but staying linear? But I can't, neither in Raw Developer nor on the linear TIF in my image editor, PhotoLine.
As soon as I lower the y-value of the white point, the sky turns muddy, whereas it is clear in the dark image.

Have you tried in ACR/LR ?
I don't know about Raw Developer but iirc Canon's DPP has such issue as well,
i.e. truly clipped '255' white is pulled down to gray.

Peter

--
Logged

jbrembat

  • Full Member
  • ***
  • Offline Offline
  • Posts: 177
Re: Problem adjusting white point in post-processing
« Reply #4 on: January 06, 2011, 06:11:27 am »

Quote
Originally, I intended to stack these. But then I discovered that BOTH needed the SAME adjustment of the white point in Raw Developer to become "clip-free" in the highlights, namely 255/254
White balance is not for clipping removal.

If you have clip the white balance can "hide" the clip, but starting values for white balance are clipped, so white balance on that colors is incorrect.

Exposure problems are to be counteracted using exposure.

If the scene contrast is so strong that camera sensor is unable to represent it, you have to switch to multiexposure HDR or stay with clipped colors.

Jacopo
Logged

Hening Bettermann

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 945
    • landshape.net
Re: Problem adjusting white point in post-processing
« Reply #5 on: January 06, 2011, 02:20:28 pm »

Hi Peter,

thanks for your reply.

I had not tried ACR, but upon your question, I did, and indeed, ACR performs as exspected.

The screen shots show the upper right corner of the image. In ACR, Exposure is set to -1 EV. In RD, the white point of the tone curve is lowered from 256 to 128. Don't mind the color difference. RD is with my home made camera ICC profile, ACR is with the ACR 4.4 profile.

« Last Edit: January 06, 2011, 02:34:57 pm by Hening Bettermann »
Logged

Hening Bettermann

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 945
    • landshape.net
Re: Problem adjusting white point in post-processing
« Reply #6 on: January 06, 2011, 02:39:17 pm »

Hi Jacopo,

thanks for your reply.

>>But then I discovered that BOTH needed the SAME adjustment of the white point in Raw Developer to become "clip-free" in the highlights, namely 255/254
>White balance is not for clipping removal.

The white point I was talking about is the top right end of the tone curve, not the white point of white balance. While the same term is used in either case, my mentioning the values 255/254 should have set the context. The "Exposure" slider in ACR does the same: move THIS white point. - I had taken 2 exposures to cover the contrast of the scene, but then I discovered, that I might do with just the light exposure.

Good light! - Hening

Hening Bettermann

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 945
    • landshape.net
Re: Problem adjusting white point in post-processing
« Reply #7 on: January 06, 2011, 02:51:12 pm »

For a moment, I thought I had found the solution. Under the In/Camera Advanced tab, RD has adjustment sliders for the black and white point. The white point was at its default for this camera, 234. But even setting it to the obtainable max, 250, makes no difference with regard to my observation.

Hening Bettermann

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 945
    • landshape.net
Re: Problem adjusting white point in post-processing
« Reply #8 on: January 07, 2011, 08:22:46 pm »

Hi again!

I asked Brian Griffith, author of Raw Developer, to have a look at this topic. He answered promptly and extensively as usual. His answer is below, preceded by his authorization to me to post it here:

"No problem posting this information. The only note I'd add is that I personally have no internal knowledge of exactly what Adobe is doing inside their products or precisely how the exposure slider functions in ACR/Lightroom, but I think my description is roughly accurate based on general observations..."

------
"I'm unable to see any of the images you are discussing on the forum, so I'm not 100% certain what you are going for, but hopefully this addresses your question?

Basically you can assume clipped values are either 1.0 or infinity. Most all image editors (RAW Developer, Capture One, Aperture and even Photoshop itself when not using ACR) assume the 1.0 convention when adjusting exposure. Therefore a clipped value at say -1EV exposure will have a value of 0.5 its original value. In 8bit for example a value clipped at 255 would become 255/2 or 128 rounded up as long as the color values are strictly linear (note that most RAWDev camera curves are not strictly linear and roll off the highlights... also note that many digital cameras may have some "head room" or baseline exposure adjustment and there may be valid image data above what is displayed as 255 at exposure of 0.0, also highlight recovery is implemented in many RAW editors including RAWDev which may alter the handling of clipped values).

There's no way to know whether a clipped value is just exactly clipped or may be 100x beyond the clip point or somewhere in between. Assuming clipped values are 1.0 allows you to preserve linear tonal relationships across the full color range when adjusting exposure, as the identical exposure correction is applied to every value, and this really is the technically proper exposure adjustment. The tradeoff for this is [that]clipped white values, that you may want to assume are infinitely white or at least very much brighter than their peer, become dark gray when adjusting exposure significantly down.

As you have found, Adobe's RAW software (ACR and Lightroom) handle exposure differently, more like assuming clipped values are infinite. The problem with assuming clipped values are infinite is that in order to not give an ugly appearance when adjusting exposure down, values just below the clipping point and some range below  cannot have the exposure value fully applied. For example assuming 8bit values again if you had a value of 254 that was not clipped and set exposure to -1, that value would become 127 and say those very bright, but not clipped pixels were from some clouds that were right up against some sky that was clipped at 255 and was assumed to be infinite. You basically would end up with very dark clouds right up against a bright maximum white sky.

So if you assume clipped white values are infinite, in order to give a pleasing appearance you basically have to use a curve or non-linear exposure adjustment. This is essentially what Adobe does with ACR and Lightroom exposure. The exposure slider isn't really a true "exposure" adjustment equally applied to all values at least when using negative values.

In RAW Developer, to give a similar result you would need to use a manually created curve where the white point is kept at the maximum or close to it  and bright highlights are rapidly rolled off (plunge downward). Over most of the image tonal range the curve would be linear say to give a -2.0EV adjustment the output values would be exactly 1/4 the input values...

Here's a visual example, not the greatest example but I had this image sitting on my desktop..., of a -2.0EV adjustment and the differences between RD and ACR visually and how a curve (very, very quickly thrown together here...) can approximate the behavior of ACR's exposure slider:"
------


« Last Edit: January 07, 2011, 08:30:13 pm by Hening Bettermann »
Logged

jeremypayne

  • Guest
Re: Problem adjusting white point in post-processing
« Reply #9 on: January 07, 2011, 09:19:10 pm »

The difference between converters is interesting.

But ... Back to your original captures ...

Don't you think you would have been served better if you had exposed the "darker" frame in such a manner that the sky was NOT clipped?  Wasn't that the point of taking multiple frames in the first place?
Logged

Peter_DL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 544
Re: Problem adjusting white point in post-processing
« Reply #10 on: January 08, 2011, 04:34:33 am »

Brian Griffith wrote >>  There's no way to know whether a clipped value is just exactly clipped or may be 100x beyond the clip point or somewhere in between. Assuming clipped values are 1.0 allows you to preserve linear tonal relationships across the full color range when adjusting exposure, as the identical exposure correction is applied to every value, and this really is the technically proper exposure adjustment. The tradeoff for this is [that]clipped white values, that you may want to assume are infinitely white or at least very much brighter than their peer, become dark gray when adjusting exposure significantly down.

As you have found, Adobe's RAW software (ACR and Lightroom) handle exposure differently, more like assuming clipped values are infinite. The problem with assuming clipped values are infinite is that in order to not give an ugly appearance when adjusting exposure down, values just below the clipping point and some range below  cannot have the exposure value fully applied. For example assuming 8bit values again if you had a value of 254 that was not clipped and set exposure to -1, that value would become 127 and say those very bright, but not clipped pixels were from some clouds that were right up against some sky that was clipped at 255 and was assumed to be infinite. You basically would end up with very dark clouds right up against a bright maximum white sky.

So if you assume clipped white values are infinite, in order to give a pleasing appearance you basically have to use a curve or non-linear exposure adjustment. This is essentially what Adobe does with ACR and Lightroom exposure. The exposure slider isn't really a true "exposure" adjustment equally applied to all values at least when using negative values.

Interesting, explains why we once found Exposure to behave not entirely linear in ACR.

Peter

--
Logged

Hening Bettermann

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 945
    • landshape.net
Re: Problem adjusting white point in post-processing
« Reply #11 on: January 08, 2011, 08:39:29 am »

Hi Jeremy,

> The difference between converters is interesting.

Yes, and I for one am happy to learn that "my" converter uses a true linear approach, now that I understand it.

> Don't you think you would have been served better if you had exposed the "darker" frame in such a manner that the sky was NOT clipped? 

Yes that was the intention but obviously I did not quite succeed.

> Wasn't that the point of taking multiple frames in the first place?

It is 1/2 of the point. The other half is to replace noise in the shadows with detail from the light exposure.

Good light!

Hening Bettermann

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 945
    • landshape.net
Re: Problem adjusting white point in post-processing
« Reply #12 on: January 08, 2011, 09:14:24 am »

Hi Peter_DL,

>Interesting, explains why we once found Exposure to behave not entirely linear in ACR.
> Yes, and I for one am happy to learn that "my" converter uses a true linear approach, now that I understand it.

And the more so if I see (again) which tech stuff I would have to digest to understand what is actually hiding under various Adobe labels.

Good light!

Peter_DL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 544
Re: Problem adjusting white point in post-processing
« Reply #13 on: January 08, 2011, 10:41:24 am »

Yes, and I for one am happy to learn that "my" converter uses a true linear approach, now that I understand it.

Hening, - so which advantage do you see which such truly linear implementation of an Exposure slider ?

I'd rather be concerned that any burned reflection spot is turned to gray when pulling down an ETTR image.

Sincerely,
Peter

--
« Last Edit: January 08, 2011, 11:10:35 am by Peter_DL »
Logged

Hening Bettermann

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 945
    • landshape.net
Re: Problem adjusting white point in post-processing
« Reply #14 on: January 08, 2011, 01:04:21 pm »

Hi Peter_DL,

I don't have the technical insight that you have and other members of this forum and the Adobe forum, but this passage from Brian Griffith's reply makes sense to me:

> Assuming clipped values are 1.0 allows you to preserve linear tonal relationships across the full color range when adjusting exposure, as the identical exposure correction is applied to every value, and this really is the technically proper exposure adjustment.

I feel like everything non-linear that occurs under the table prior to the final TRC is something that goes on out of my control. This may be naïve, since a lot of things probably go on under the table that I don't know of and would not understand. But I try to stay in control as much as possible.

Sure, as I have experienced, ACR is easier to handle on this point, but the concept of theoretical correctness appeals to me.

Good light - Hening.

Peter_DL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 544
Re: Problem adjusting white point in post-processing
« Reply #15 on: January 09, 2011, 11:45:19 am »

I feel like everything non-linear that occurs under the table prior to the final TRC is something that goes on out of my control. This may be naïve, since a lot of things probably go on under the table that I don't know of and would not understand. ... the concept of theoretical correctness appeals to me.

Hening, - the idea of a truly linear Exposure slider may sound appealing,
however, with the images I have in mind it is probably better to leave truly clipped white at infinite '255' rather than risking to turn it to gray.

It was stated that this can also be reached by a curve in Raw Developer, however as far as I can tell ACR seems to be differently construed. Maybe the crux is Highlight recovery: ACR attemps to reconstruct useful luminosity information when only one or two channels are clipped. Hence, it can't be truly linear anymore.

It would be interesting to listen a related discussion between Brian Griffith and an ACR engineer.

Peter

--
« Last Edit: January 09, 2011, 11:48:28 am by Peter_DL »
Logged

Hening Bettermann

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 945
    • landshape.net
Re: Problem adjusting white point in post-processing
« Reply #16 on: January 09, 2011, 12:57:22 pm »

Hi Peter,

I'm not sure if higlight recovery has anything to do with this. Both converters have this, and it can be turned on and off. (Was off in my case).

It could indeed be interesting to hear the experts talk to each other (as we often do on this forum, don't we?).

For my own part, my conclusion so far is:
1- I am happy that now I at least understand why I had problems and what they are about.
2- If I find constructing the curve in RD too difficult, I have 2 choices:
 a- stack the images, if I have different exposures
 b- if not, I could develop linear in RD and then edit the Exposure on the .tif in ACR, before I fine tune the TRC in PhotoLine using a luminosity mask.

Good light! - Hening.
Pages: [1]   Go Up