Jump to content
Brad Brighton

Tone Mapping Madness?

Recommended Posts

So, I have a panorama I'm working on comprised of HDR-merged components. I freely admit that I hack some of this stuff as I'm learning when it comes to which adjustments lead to what results, as long as I like the visual endpoint. That being said, this tone mapping interim result (attached) was a bit unexpected. The screenshot was with 1.7.0.116 (beta) but the reason I was there was because I saw the issue in production 1.6.11 and wanted to confirm before submitting the question (so this is not a beta-specific issue). I've also attached a 1.6.11 screenshot pre-entry into the Tone Mapping persona for comparison.

Is this a bug? Is it broken-as-expected? Is this somehow user error?

I can privately submit the project and/or the exrs if needed.

Thanks!

-brad

 

Screen Shot 2019-04-04 at 9.21.02 AM.png

Screen Shot 2019-04-04 at 9.25.07 AM.png


https://bmb.photos Focus: The unexpected, the abstract, the extreme, including paper outputTools: macOS (Primary: Mojave, MBP2018), Canon (Primary: 5D3), iPhone (Primary: X), Epson

Share this post


Link to post
Share on other sites
Posted (edited)
7 hours ago, GabrielM said:

Hi @Brad Brighton,

I would say it's not expected. Can you please upload the resources here? https://www.dropbox.com/request/mpEAprPZAhZLzCvrGQMF

Thanks,

Gabe. 

I'll be dropping rather large source materials shortly. As I'm working out exactly which sources to send, I'm suspecting that the panorama render result may be playing an unexpected role -- I'm seeing what appear to be some unfriendly changes to the sources once the merge is complete, at least as compared to inspecting the individual EXRs. The weirdness may not be visually apparent before attempting to tone map...

The pano is comprised of a single "aces" EXR, a "Harrahs" EXR, and a "Circus" EXR. The variants of Aces (and I have several more available, in case they are needed) were an attempt to get rid of the resultant cast in the merged panorama that doesn't seem to appear (or at least is not prevalent) in the individual EXRs.

The lens flare is known and not knowingly part of this issue.

System:

MacBook Pro 2018 i9/32G/Vega20

Mojave 10.14.4

Original images: Canon 5D Mark III Raw (CR2)

If there's anything else you need, just shout!

Edited by Brad Brighton
Additional information

https://bmb.photos Focus: The unexpected, the abstract, the extreme, including paper outputTools: macOS (Primary: Mojave, MBP2018), Canon (Primary: 5D3), iPhone (Primary: X), Epson

Share this post


Link to post
Share on other sites
7 hours ago, GabrielM said:

Thanks. We're currently looking into this. Will let you know when we have an update :)

Thanks much! 


https://bmb.photos Focus: The unexpected, the abstract, the extreme, including paper outputTools: macOS (Primary: Mojave, MBP2018), Canon (Primary: 5D3), iPhone (Primary: X), Epson

Share this post


Link to post
Share on other sites
Posted (edited)

@GabrielM I have another unexpected Tone Mapping case that may be more useful as it's far less complicated. 

A CR2 raw image single image (that's meant to be part of an HDR merge, hence the low exposure values).

Whether I use 1.6 or 1.7 beta, whether I use the out-of-camera original or the DPP4-processed version to attempt to remove luminance noise, I get a default tone map result like the attached screenshot. 

I don't actually want to tone-map this image yet in my workflow but I think this may highlight (pun intended) something either wrong with my processing flow (user error?) or with the processing itself.

The attached ZIP file holds the OOC original, the save-history-enabled .aphoto, and the DPP4-processed original. In my testing, both of the originals generate an effect like this whether any adjustments have been applied or not.

Repro steps:

  • Open one of the provided files - if it's a CR2 file, Develop without changes.
  • Enter Tone Mapping persona
  • Profit!

Nothing has changed in the system config since the previous report. The Canon 5D Mark 3 firmware *has* been updated to 1.3.5 and the lens registration information for the EF 75-300 III hs been installed. I'd say there's a difference in detail but not in action between the Serif RAW and the Apple RAW processing, though that in itself may be a clue. Develop assistant settings also attached (and the incentive to try the Apple RAW processing as well).

[EDIT] Exporting to 16- or 8-bit TIFF and re-editing does not make this effect go away. It does, of course, change the specifics of the result, still washed in the highly pink haze though.

Screen Shot 2019-04-18 at 10.19.29 AM.png

Screen Shot 2019-04-18 at 9.50.03 AM.png

for_affinity.zip

Screen Shot 2019-04-18 at 10.16.20 AM.png

Edited by Brad Brighton
Added the details about the experimental TIFF variant exports.

https://bmb.photos Focus: The unexpected, the abstract, the extreme, including paper outputTools: macOS (Primary: Mojave, MBP2018), Canon (Primary: 5D3), iPhone (Primary: X), Epson

Share this post


Link to post
Share on other sites

This is expected with your sample. Even if you just increase the exposure, you will have the same effect. All that pink noise is already present in the image, but to some very-low values. When you increase exposure, it becomes more noticeable. Tone-mapping an image that's nearly black is not a good idea, as there is no data to recover from it. 

Share this post


Link to post
Share on other sites
9 hours ago, GabrielM said:

This is expected with your sample. Even if you just increase the exposure, you will have the same effect. All that pink noise is already present in the image, but to some very-low values. When you increase exposure, it becomes more noticeable. Tone-mapping an image that's nearly black is not a good idea, as there is no data to recover from it. 

@GabrielM (and anyone else who has something useful to add),

Well, hmmm...

  1. That's probably the issue with the original post in this thread then -- these nighttime RAW shots have a fair amount of low-level noise in them and the original is lots of these put together using HDR merge and panorama merge. (Also see #2)
  2. Apparently, the Canon 5D series exhibits lots of noise (luminance) in extreme low-and-high light situations like this (I found many, many complaints about this specific situation "on the web", salt-grain taken)
  3. I can't seem to remove this noise with either production or beta Photo (even allowing for DPP4 pre-processing)  -- I'll chalk this up to operator error but assistance in getting to a solution using APh would be greatly appreciated since the "noise" gets magnified unacceptably whenever I either HDR merge or Panorama-merge (even before tone mapping) and if I try to tone map the result, all color hell breaks loose.
  4. Maybe (probably) there's something fundamental I'm missing WRT to using HDR and/or tone mapping to bring out shadow detail in nighttime shots such as this. Where should I look to correct this lack of knowledge? I am used to the tools at hand being able to manipulate camera- and shooter- errors (to a point, of course) but this specific scenario seems completely outside the current APh wheelhouse based on your response.

Help?


https://bmb.photos Focus: The unexpected, the abstract, the extreme, including paper outputTools: macOS (Primary: Mojave, MBP2018), Canon (Primary: 5D3), iPhone (Primary: X), Epson

Share this post


Link to post
Share on other sites

I don't believe there's much you can do about that noise. There is barely any information in the sky, so the only info that you can "recover" is the noise. 

14 hours ago, Brad Brighton said:

Maybe (probably) there's something fundamental I'm missing WRT to using HDR and/or tone mapping to bring out shadow detail in nighttime shots such as this.

HDR is mainly used during the day, where you can really push the values. At night, you barely have any information to push. I suggest you look at long exposure photos. Perhaps one for the sky, one for the city and one for the foreground. You can then try to stack them and see what you come up with. 

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×