HDR - Open EXR-Output

[Volker Harun]

Just for preventing me to spam another thread, I set my question here.

The last weeks, I saved my images as OpenEXR, to use the HDR features of Photomatrix of PS CS2.
It is nice, that the sky's lighting is enhanced, but the shadows still remain boring.
Is the OpenEXR Data calculated by GI?

Thanks,
Volker

[Matt]

Everything in the renderer is calculated in high dyanamic range, whether you output to EXR or a low dynamic range format such as BMP. The only difference is that the EXR image preserves most of that high dynamic range without having to clip it or quantize it so much. The EXR is just a more accurate record of the pixels on the screen - it has nothing to do with GI in particular. If the shadows don't look very dark in the BMP then the EXR is not going to magically create more contrast.

If there is not enough detail/contrast in the shadows then that depends on the lighting conditions, surface textures and GI settings (GI relative detail, GI sample quality, GI surface details). For a given example scene I can probably offer advice on achieving more interesting shadows, but every scene is different.

Matt

[Volker Harun]

... the EXR image preserves most of that high dynamic range without having to clip it or quantize it so much.

Hi Matt,
thanks for the response.
Does the above mean that it will be more accurate when making 2 or 3 exposures of one scene and to combine them in the editor of choice?
For example, with my camera's raw-output I can create a pseude-hdr, it can be tweaked by some means. But for a pure HDR I need 3 or more exposures: HDR-example

Is it like that?

,-) Volker

[Matt]

No, with computer generated HDR there is no need to render multiple exposures. That is only necessary with photography because the camera is a low dynamic range device. CG is not bound by such restrictions

Matt

[Volker Harun]

I'll go for some example scenes - soon ,-)

[Matt]

Another way to put it is this: with a camera you can usually only capture a limited range of intensities, so if you want to work with a higher dynamic range you have to reconstruct an estimate of that information somehow. With a high dynamic range renderer (most are these days), the renderer created that information in the first place so it can easily write it straight to a high dynamic range file (eg. .HDR or .EXR) in a single render.

Matt

[Sethren]

Speaking of HDRI output. Might there someday be a possibility that Terragen could render a full spherical panorama scene? I can do this in Carrara 5 and Vue 6 but i'd very much like to try this in Terragen 2. Altough Carrara does not have any means of HDRI export so i have to render 2-3 scenes at different exposure levels. Vue 6 is Vue 6, a crash ridden disaster and a very slow renderer. A part of my work flow involves HDRI Light Probes but the scenes need to look as real as possible and i know TG2 will take care of that part of it. I know there are scripts that can do this but it would be very easy to just render that single output and then save it as an HDRI image.

[Volker Harun]

I like to be proven wrong - at least in terms of underestimating the capabilities of this program.
Attached is one image that is showing the luminance of a picture's shadows and one shows this render's highlights.
Chapeau!

[bigben]

Speaking of HDRI output. Might there someday be a possibility that Terragen could render a full spherical panorama scene? I can do this in Carrara 5 and Vue 6 but i'd very much like to try this in Terragen 2. Altough Carrara does not have any means of HDRI export so i have to render 2-3 scenes at different exposure levels. Vue 6 is Vue 6, a crash ridden disaster and a very slow renderer. A part of my work flow involves HDRI Light Probes but the scenes need to look as real as possible and i know TG2 will take care of that part of it. I know there are scripts that can do this but it would be very easy to just render that single output and then save it as an HDRI image.

You can render HDRI cube faces and then convert them to a HDRI panorama.  I've done this with TG0.9 and it should be no different with TG2 as the conversion process is independent of the rendering application. If you have the animation version, you can script the cube face creation.  If there are plans to include equirectangular renders I dare say it would be a fair way down the track.

Just for preventing me to spam another thread, I set my question here.

The last weeks, I saved my images as OpenEXR, to use the HDR features of Photomatrix of PS CS2.
It is nice, that the sky's lighting is enhanced, but the shadows still remain boring.
Is the OpenEXR Data calculated by GI?

Thanks,
Volker

Hi Volker

Here are some screen grabs of one of my HDRI conversions that will hopefully clarify some things for you. I convert to 16 bit initially just in case I want to tweak the colours a bit more afterwards.  The default settings that Photoshop chooses for the conversion from HDRI depend in part on the histogram of the image. If you have a particularly high key or low key image then this will probably look awful (image 1 has blown out highlights). Either way, you have full control over the conversion process by editing the curve on the histogram. You can start by moving the black and white points, giving you a linear contrast (3rd image), but I usually end up making a curve that looks more like a photographic film's characteristic curve (last image) which IMO improves the perception of realism.

[Sethren]

I know about the scripts and cube faces. I was hoping Matt would have some type of an answer. I do appreciate the help though. I will probably try that in the final release and when i have a faster liquid cooled system.

[Saurav]

It's always better to apply post processing at higher bit rate (higher dynamic range) if you can, this prevents issues such as banding you see sometimes in TG images. But take into account that even Photoshop CS3 only supports limited 32bit image filters. There are other tools to work with HDR images, but I haven't had much experience with them.

The usual work flow for online posts is as follows for me:

1.EXR render import into Photoshop CS3.
2.Adjust exposure, levels, hue saturation, sharpness, DOF (if necessary) in 32bit mode.
3.Convert to 16bit.
4.Colour balance
5.Export to relevant format.

[Volker Harun]

I think that all I need is a different approach to the image.
I.E. in photography I go for 'underexposed' images, as I do like the contrast of shadows, but I want the sky to become dark.
I use exposure settings of -4|-1|+1.
Then the postprocessing is fun.

In TG2 I went in the preview for a well exposed image, so I got a flat image no matter how I did the postprocessing.
I just have to be a bit more thoughtful, to get that HDR that I am expecting.
Another point are clouds - they turn out to be very flat after conversion - this will be for future renders my measurement to produce good cloud shaders ,-)

Volker

[bigben]

If you use a curve like my last example above this should do what you want.  With HDRI you get far more highlight detail, so when you do this kind of conversion it pushes the sky's tone back towards the mid-tones, so a curved adjustment like this will darken the sky while you are tweaking the shadow's "exposure"

I haven't done a HDRI image with thick clouds yet, but it may require some rethinking of the cloud's lighting settings. If you're going for dense clouds you might expect to see some blowout of the highlights in the preview.

You might also try playing with PS's Shadow/Highlight adjustment after converting the HDRI to 16 bit. Unsharp masking can be more effective sometimes as an interpretive adjustment.

[Volker Harun]

Hi Bigben - in my HDRs I have at least two peaks, these are quite hard to handle ,-) But I got your point.

I used to use the Shadow-Highlight-Adjustmenst in PS, but today I use calculations for masking areas of interest and do the adjustments manually - at least for photographies.
For renders I normally just frame, maybe a tit of softness but not very often.

Volker