[NOOBTUT#2] Tonemapping with 32-bit Color Depths

Created a nice video, document or a translation of another document? Share it here for licensed users.
User avatar
JQL
Thea Beta Tester
Thea Beta Tester
Posts: 5311
Joined: Sun Jun 23, 2013 6:34 pm
Location: Lisboa
Has thanked: 22 times
Been thanked: 118 times
Contact:

Last [NOOBTUT] was all theory and no practice. We've seen the difference between 32-bit and 8-bit and what each one is good for.

This time we'll get our hands dirty.


Disclaimer:

Still a lot of trial and error aproach in this Noob quest of mine, so there will be still a lot of imprecise stuff in here...



32-bit what to do with all those colors?

What you want is to use those almost 5 billion colors and show all of the details that are hidden away. We do that by filtering out a lot of colors that you don't need, lowering the high values or raising the low values of a pixel color, in order to make it fit inside an 8-bit color depth image, thus becoming visible.

In 8-bit terms what we are talking about is turning brighter than white pixels into white or lower colors. Or turning darker than black pixels into black or higher than black colors. For this to happen we have to sacrifice something that is already inside the 8-bit. This, of course are mid range pixel colors. This way we can show some of the objects hidden in the dark or the light by degrading the colors of the objects inbetween, in a process that can be seen as a selective degradation of the image from 32-bit to 8-bit.

Opposite from what we've seen, the dangers of Tonemapping with a 32-bit image are less and just have to do with color itself as all pixel detail will be there all the time and you won't easily get that flat pixel colors that happened in the TRY IT section of the previous TUT.

So In the end, what you do is up to you, but usually two kinds of things are being done - emulating a camera or emulating the eye/brain produced image.

EXR 9 Different tonemapping Brightness, Saturation, Contrast.jpg

1 - Emulating a Camera Film/Sensor response:

In short terms what a real camera does is either focus on Dark areas or Bright areas of an image, excluding almost all details from the other, or focusing on the midrange and exclude some details of both high and low ranges (bright and dark areas/colors).

The reason photorealistic rendering is doing what it does by emulating cameras is that, for a century, our culture has been developed around the idea that all 2D image output from cameras is the most realistic that we can get.

The idea that you can render a 3D to create such a realistic output that you can convince someone that you shot it with a camera is all that's behind the concept of Photorealism and it is basically the mantra of most rendered production these days.

Physically Based Rendering and Photorealism are two different things though. The first has to do with the process of light calculation, the later has to do with how you show light calculation.

Photorealism is very successful, partly because people are not used to seeing images of reality that look the same as the way we actually see reality with our own eyes. So, to make a render look Photorealistic you actually have to make it different from what you see in real life, and emulate the limitations of a camera.

Thea darkroom is all about that!

We've reached a point of film/camera limitations emulation that you nowadays have to sacrifice render times to distort a raw render with effects like, Depth of Field, Motion Blur, Exposure settings, Vignetting, Glare and Film Response, etc. Particularly film response is very interesting because old films or old photo paper distorted light values so much, and there were so many kinds of film and paper, that we, nowadays use that as retro or mood capturing effects. Instagram does it and even Thea does it!

It's the CRF part of your Darkroom...

So, if you aim at emulating most of what a real camera does or used to do, simply head to this excellent tutorial by Patrick Nieborg, who is Thea's Technical Artist and has a keen eye for everything light and material related:

https://www.thearender.com/resources/Ac ... posure.pdf


2 - Emulating the human eye/mind

Where the human eye and mind differ from a camera is that they can perceive darker and lighter parts of reality in a much wider range than a camera film/sensor. They have many of the same limitations a camera has so all those distortion effects are also used to emulate the way humans see reality. The difference from a camera though is mostly that all things work better with us than with cameras. Our eyes have wider lenses, our retina is a way more sophisticated sensor device, but mainly our brain is a much more advanced hardware and our mind can associate information selectively much faster than any software or algorithm can.

Also, recent years have seen the development of a photo technic called High Dynamic Range (HDR). This photo technic is based on post production, where photographers overlap several shots with varying exposures, each exposure focusing the image on specific intensities of the dark, middle and bright ranges of colors. After all have been blended, the resulting file might be no more no less than a 32-bit image. This method is also widely used for your IBL images as that 32-bit depth images carry whiter whites, thus stronger and precise lighting, that enable it to be used in your IBL with great quality.

The most common file formats for these 32-bit images are HDR and EXR but Thea renders are basically High Dynamic Range images too.

The trouble here, is again that there is no medium that can fully display an HDR image so it's you that have to do all the selection work until you find what looks right and export it to 8-bit.


3 - Expand known boundaries

There are of course no barriers to bind use and we could use many ways of pushing a 32-bit image to all it can get!

I discussed this a bit with Patrick Nieborg of interpret what he said, I guess it'll just be better if I quote him:
Patricks wrote:... one needs to understand that 80% of the "wow factor" seen in movies and Magazine images are achieved in post pro. If you would see the raw footage of any movie or the raw photo of any magazine, you would not believe that its the same image used in the final product.

It's like with the Hollywood actors and music stars or top models... they all get their Photoshop treatment ... and for any other image it's not different.

Reality is not "beautiful" ... its the interpretation we make as artists that makes it beautiful and this is made through image manipulation/post pro.

Sometimes as photographers we can capture a special moment and the raw image is the final one... but this is a very rare case and 99% of photos (even in the old days with chemical films) have some kind of manipulation/post pro.
With that in mind let's start...



32-Bit Formats EXR vs IMG.THEA


The first thing you have to do is to save the image as a 32-bit depth IMG.THEA, HDR, TIFF or EXR for later tonemapping.

I've purposedly introduced IMG.THEA format here as it is one of the best 32-bit formats you have available for use with Thea.

Basically, if you're working towards camera emulation, as we've seen previously you can do everything in Thea's Darkroom and what you're doing is tonemapping thea's native 32-bit depth format.

However, this format is constrained to Thea, and we want to broaden our scope here, so we'll choose one of the other popular formats.

I'd recommend EXR over HDR or TIFF for this. It's lightweight (sort of...), it supports layers, and it's an open format, so it's becoming standard for 16-bit or 32-bit workflows.

Let's do it then!

1 - Saving an EXR:

Now, since Thea version 1.5 has come along, we can export multilayered EXR files with all channels we choose to render. For a beginner this might not mean much, but as soon as you progress in your expertise, this will be an unvaluable asset for your post production. Channels and multilayered EXR, however it's a story for another time.

For now the only channel you'll need is the Color Channel. So, from Thea Studio 1.5, you'll click the save image button:

- You'll have to choose the format to EXR
- Thea 1.5 will show up a table;
- If you've activate the multilayered EXR export, you should deactivate it for now and it will export only the color channel.
- Set the color depth to 32-bit on the color channel.
- Name the image and hit save.

If you didn't export your EXR yet, you can use this one:


EXR Tutorial Base.zip
(3.68 MiB) Downloaded 47 times


I think it's interesting to talk a bit of this image. It had an original resolution of 1620x2160 but it's EXR size was of around 48Mb, wich is too big for sharing here.

So I shrunk it down to 500x666 until it had a total size of around 3800Mb. Some details are lost, but who cares about details anyway when you have 32-bit color depth, right?

Actually the opposite is the better truth. If you're going to work with 32-bit then it's all about details. The 32-bit formats will allow to see all of the details, very clearly. With 32-bit your quality will become pro level. But I don't do [PROTUTS] that's Patrick! So let's keep this noob profile of ours.

SIDENOTE:

As you can see this 48mb size is way more than a JPG or PNG could be for the same resolution... and you should think EXR as a lightweight format! There are heavy weight formats and in the extreme of that line, the heavier images we are usually saving on our computers are IMG.THEA files. The IMG.THEA file for this original image (1620x2160) is about 1Gb. So 1Gb is the amount of info a single thea render generates. In this same project (since I got my Titan X) I've generated renders with greater resolutions and I've reached IMG.THEA files of 3.8Gb... Of course you shouldn't directly compare EXR with IMG.THEA files because even if they are both 32-bit format images, IMG.THEA files carry much more info than just pixels, they are like a frozen state of a render project and if you want you can open one of these files and the respective SCENE.PACK and, for instance, keep on rendering it for higher samples.


But let's focus on EXR again.

You got an EXR already. Let's Tonemap it!

As we've seen already, the idea here is that you can choose if you want to focus on the brighter pixels, the darker pixels or medium graded pixels.

What EXR can do for you that you is that you can enter What You See Is What You Get mode (WYSIWYG) without the camera based restrictions Thea imposes on you, as they can only be openned by Thea inside Darkroom.

At a base level, we are talking about black and white levels (or just levels), brigtness curves (or just curves) and other color/brightness manipulators like Contrast and Saturation.

Mainly we'll focus on Levels and Curves, but most of the time you'll need contrast and saturation too, to get best results.


32-bit depth colors editing Software?

There are several, from the simple, to the very complex, from the free to the very expensive.

If what you want to do is, simply put, Tone mapping, without any photo editing or retouching you can download Luminance HDR

or Picturenaut. This is what will do for now!


Both of these are free but I'd recommend Luminance HDR over Picturenaut as it has a very good choice of presets and all of them can be finetuned. Besides that, you have a Levels and White Balance buttons, wich you can use very fast to get good results.


Luminance HDR
Luminance UI.jpg
With Luminance usage is very straight forward:

When you open an EXR file a central preview from it will popup. In the case of the EXR I've provided it will be very dark. This is exactly the same tonemap I was seeing in Thea Studio.

In the main UI you'll see 3 main areas: Main preview, preset settings to the left and preset thumbnails to the right.

The fastest way to start your tonemapping is from the thumbanail presets on the right:
1 - Simply choose the one wich you like best and click on it:
2 - A new tab is created with an 8-bit image (LDR) at a lower resolution. This low resolution has to do with the time it would take to process big IBL images like the ones used to Illuminate your scenes.
3 - If you're happy with the default tonemap you got, you can increase resolution right above "Tonemap" button on the left pane.
4 - Apply final output resoultion and click tonemap.
5 - A new image with same tonemap is created with correct resolution, replacing this one.
6 - You can save this image as a JPG.

NOTE: After you changed resolution once, every operation from then on, will keep that new resolution.

You can also finetune your tonemapping or start by selecting the tonemap type you want instead of using thumbnails:

1 - Be sure you're in the tab you want to edit, or select the EXR tab in the top if you want to try a new tonemapped version of

that EXR.
2 - On the left panel select the type of tonemap you want with the first dropdown menu
3 - You can tinker with the settings;
4 - For you to see the effects you have to click in the Tonemap button.
5 - From the moment you do that, you start working on the new LDR tab that was automatically created or you keep working on the same tab if you were editing a LDR.

If you're happy with the detail and colors you got and but if you still think it's to Bright or Dark, you can use Levels:

1 - To use Levels click on the levels button on top and tinker with the sliders as explained in [NOOBTUT#1]

If you feel overall the white level is wrong, (for example too yellowish or too bluish), you could click the white balance button on top. I prefer doing that in other software if I have too, as this is fully automatic with no user control.


CONCLUSION

As you'll probably be aware after testing all presets, there's a multitude of presets to choose from and these are able to represent very well how you can push up a 32-bit image. These kind of presets deal with all kinds of different things:

- Luminance levels;
- Detail enhancement;
- Contrast;
- Sharpenning;
- Color correction or enhancement;
- Etc...

There is a lot of software that can fake those effects from an LDR image. However if used with purpose, there's no limit to what you can control from a nice resolution, 32-bit image.


NEXT [NOOBTUT]

Next time we're going to see how to do this kind of stuff in other kind of software:

- Photo editors, wich have a lot of benefits besides tonemapping;
- Image/video compositors wich offer different workflows (like node based) and are focused on nondestructive editing of an image.

Say your mind:

Now that you got your hands on Luminance HDR, I hope you can have a better feeling of what you can achieve and again, I urge you to use the reply section to ask stuff and share your own knowledge of the subject.

It would also be cool if you'd show what you did with the provided EXR. How hard did you push it? Did you destroy it?


All the best and see you soon!
www.casca.pt
Visit us on facebook!

Win 10 64bit
Sketchup 2017 64bit
Intel i7 5820K 3.3Ghz 32Gb Ram
Nvidia Titan X 12Gb
User avatar
gwengosch
Licensed User v1.5
Licensed User v1.5
Posts: 907
Joined: Sun Jun 09, 2013 6:44 pm
Location: france
Has thanked: 6 times
Been thanked: 12 times

It's a very good thread, i wish all the tutorial would be like this one.
Thanks
Gwenael Gosch
La fabrique à plans
http://www.lafabriqueaplans.fr
Facebook
User avatar
JQL
Thea Beta Tester
Thea Beta Tester
Posts: 5311
Joined: Sun Jun 23, 2013 6:34 pm
Location: Lisboa
Has thanked: 22 times
Been thanked: 118 times
Contact:

Thanks Gwengosh!
www.casca.pt
Visit us on facebook!

Win 10 64bit
Sketchup 2017 64bit
Intel i7 5820K 3.3Ghz 32Gb Ram
Nvidia Titan X 12Gb
User avatar
archplanviz
Licensed User v1.5
Licensed User v1.5
Posts: 544
Joined: Tue Sep 10, 2013 8:52 am
Location: Austria
Has thanked: 0
Been thanked: 0
Contact:

Great JOB!
Definetly interested in the upcoming parts!

It is NEVER a bad idea saving your renders as img.thea!
Thea Render Reseller and certified Trainer

Thea Render Schulungen

Like me on Facebook
DavePSB
Licensed User v1.5
Licensed User v1.5
Posts: 1291
Joined: Wed May 12, 2010 8:56 pm
Location: France
Has thanked: 13 times
Been thanked: 16 times

Thanks for taking the time and effort to put something like this together. It really is a great contribution to the community, and worth the long hours it must have taken to prepare. I have played with Luminance HDR before but not got very good results. I will have to try it again now!
Win 10
i7-9700K,3.6GHz,32Gb,
RTX2060Super, 8Gb, 442.19
SUMake2015 (15.3.331)
TheaForSU (1.5.09.547.1462)
Thea Studio (1.5.09.1462)
SUMake 2017 (17.2.2555)
TheaForSU (2.2.954.1860)
Blender 2.79, Plugin 1.5.8.760.1455
Blender 2.8x
Revised 14 Feb 2020
bizango
Licensed User v1.5
Licensed User v1.5
Posts: 8
Joined: Mon Jul 15, 2013 10:05 pm
Has thanked: 0
Been thanked: 0

Great tutorials!! You just answered about 28407082 questions I've had for a looooong time!

Thank you :D
User avatar
Bob James
Licensed User v1.5
Licensed User v1.5
Posts: 1482
Joined: Sat Nov 21, 2009 6:57 am
Location: Lompoc, CA USA
Has thanked: 4 times
Been thanked: 2 times

DavePSB wrote:Thanks for taking the time and effort to put something like this together. It really is a great contribution to the community, and worth the long hours it must have taken to prepare. I have played with Luminance HDR before but not got very good results. I will have to try it again now!
+1
Great tutorials = Thank you
i7-6800K 3.4Ghz, GTX 1080ti 11GB, 32GB Crucial DDR4-2400 ECC, OCZ Vertex 4 1TB, WD Black 3TB, 32TB NAS, 4x Dell 27" Monitors, SpaceMouse Pro
Never underestimate the intelligence of your audience nor overestimate your own - Bob James
User avatar
JQL
Thea Beta Tester
Thea Beta Tester
Posts: 5311
Joined: Sun Jun 23, 2013 6:34 pm
Location: Lisboa
Has thanked: 22 times
Been thanked: 118 times
Contact:

Thanks guys,

It was very interesting to put these together but knowing it has reached you is really worth it.

I hope the next one opens up perspective to what you really can do if you ditch traditional post pro (photoshop style) and embrace node based or other kinds of undestructive image editing.

Please ask questions if you have them. Who know where they'll lead to,

Thanks again,

João
www.casca.pt
Visit us on facebook!

Win 10 64bit
Sketchup 2017 64bit
Intel i7 5820K 3.3Ghz 32Gb Ram
Nvidia Titan X 12Gb
jiminy-billy-bob
Licensed User v1.5
Licensed User v1.5
Posts: 304
Joined: Mon Sep 23, 2013 3:13 pm
Has thanked: 3 times
Been thanked: 7 times
Contact:

JQL wrote:traditional post pro (photoshop style) and embrace node based or other kinds of undestructive image editing.
I do all my post pro in Photoshop, and it's non-destructive. I don't see the correlation?
User avatar
JQL
Thea Beta Tester
Thea Beta Tester
Posts: 5311
Joined: Sun Jun 23, 2013 6:34 pm
Location: Lisboa
Has thanked: 22 times
Been thanked: 118 times
Contact:

jiminy-billy-bob wrote: I do all my post pro in Photoshop, and it's non-destructive. I don't see the correlation?
You can and should work non-destructively with Photoshop (Krita is also very good at that). However that is a workflow that requires a specific procedure or you will eventually affect original pixels. Node based editors are non-destructive by nature.

Node based editors let you even do other stuff that make them really worthy, but I'm just diving into them, so I will probably revise my assumptions along the way...

Thanks for your input Jiminy, I wish I had more like that!
www.casca.pt
Visit us on facebook!

Win 10 64bit
Sketchup 2017 64bit
Intel i7 5820K 3.3Ghz 32Gb Ram
Nvidia Titan X 12Gb
Post Reply

Return to “Tutorials & Translations”