AutoExposure for Vraycam

[RErender]

I'm toying with the idea of auto-exposure for the vray cam. I haven't come up with a great way to do it yet though.

I suppose it will work to sample a light value somwhere in the scene and adjust the f-stop and duration accordingly. I am stumped on where and when to sample the light values. It might work to sample the LC or IMAP for a mean brightness value and apply the exosure to the final image. Perhaps in many scenes it would work to just sample some the direct light values and go from there, ofcourse hat is dependent on having direct lights in the scene.

I found this http://www.imageval.com/public/Produ...toExposure.htm math looks like it could work.

Anyways , If you have any thoughts on it, I think it is something we could all benifit from.

[vance3d]

Along those lines it might be cool to develop a light meter helper object that you can move around in the scene and check the light levels taking into account the film rating and shutter speed.

Good Luck!

V Miller

[Da_elf]

there is a light meter plugin ive seen somewhere. we need one designed for vray now. something that can even do minute renderings based on its FOV and calculate bounced light as well. so the light meter will be a camera that analises the brightness of what its rendering.

[studioDIM]

Working Proof of concept

http://www.scriptspot.com/lele/exposure_testbed.rar

Load the scene, select the camera.
It's a default Physcam (as per creation parameters), and the scene contains a vraysun and sky, at their defaults as well.
As i said, have the camera selected (that is what you'd want to optimise) and run the script.

When you'll see a tiny 128x128 render of the scene, that's when the script thinks it's exposed properly.

So far is REALLY just a proof of concept (and it does work: sun, sky, 3 bounces.), so many things aren't in.
Notably, to be this quick, the renderer needs some peculiar setting (the ones present in the testbed scene, accidentally).
Later on the script'll take care of changing and restoring the settings at need.
Just to hammer it through: have the physcam selected.

I change the ISOs for two good reasons: F-Numbers change the DoF ranges, Shutter Speed changes the moblur lenght, but the ISOs do not change the grain size of the silver salts.
The second reason being that Shutter speed isn't common across the cameras.

Have fun with it, but don't thrash me too much, and do NOT use anywhere near a production shot.

Lele

[vance3d]

Da_Elf,

I've checked out the photometer plugin and it doesn't really fit the bill for the type of light meter I think we could use. I was thinking maybe something that would take into account direct light and maybe one bounce for a selectable number of points that could be sampled across a difinable radius. I keep wanting to have something like a sekonic.

V Miller

[Da_elf]

thats what i mean, it does a test render of what it can see then takes an average. if setup correctly you could get incident reading as well as direct readings. it will need to do a quick GI render though to be able to get the real readings with bounced light etc.

[studioDIM]

Da_elf, that's what the script IS doing, only it does it on the whole scene.
It's not nearly as accurate as a real LightMeter, but try it or take my word for it,the script WILL take you from a default, unexposed physcam, under any lighting conditions (provided for now you keep the render settings of the testbed scene) to the proper exposure ballpark.
I'm rendering an inside/outside animation to showcase how the script exposes automatically as the lighting conditions (80% of the screen is indirect lighting) vary.

Lele

[studioDIM]

Here's the sample.
Starting from default damera andvray sun and sky, i ran the script in three places i thought needed exposure adjustment, and let max interpolate.
Material is standard 128 grey.

http://www.emanuelelecchi.com/crap.htm

Lele

[studioDIM]

http://www.scriptspot.com/lele/physc...htmeter_02.mse
In OT there's an explanation of what the parameters are, and how to use it.
I'd like to understand what other things should this type of tool incorporate, if anyone has ideas.

Lele

[RErender]

The mov looks very promising nice work.

[Dschaga]

maybe this is also interessting:
http://www.kenrockwell.com/nikon/matrix01.htm

[studioDIM]

http://www.scriptspot.com/lele/Physc...eter.V1.1b.rar

Updated as per the OT topic.
DSchaga, hopefully soon i'll be able to implement something similar to that, adding some form of importance sampling to the individual pixels used to evaluate the exposure.
This said, i hope this meter works more reliably than that of current digital SLR cameras...

Lele

Edit:1.1b for bugfix. Would expose properly and then throw up an error. Didn't notice until later tonight. sorry for the inconvenience.

[werticus]

This is a very old thread now, and I was wondering if anything came of it?

Hyperfocal Design is working on a time lapse animated 360degree HDRI sky map, and an auto exposure system for the vray camera would be needed to mesh the two together perfectly.