Floating point images from MentalRay to Combustion

[irte]

I am currently in the process of testing the possible creation of a pipeline to render floating-point precision images in Mental Ray for Maya and import them into Combustion 2 for compositing. In Mental Ray options I noticed many available output formats each one with several possible channel and bit depth structures. I am slightly more interested in TIFF format use possibly, but I am open to any other option, as long as Combustion 2 is be able to handle the final rendered images with the color + alpha channels.

Any ideas on how to realize a similar combination?

I also investigated the possible use of ILM's OpenEX format (www.openex.org), but apparently the developers of that format offer only the base libraries to handle it: a Mental Ray export and a Combustion import plugin would still need to be written using those libraries. I keep this as a last resource, mainly depending on the fact I have no coder available now with the necessary Maya and Combustion API knowledge (subtle hint for anyone who would like to
volunteer )

Thank you for any info you could offer to clarify this situation.
Boonitz

[dlm]

Hi,

Combustion 2.1 can accept 32bit TGA files with Alphas.

Your other options (which don't support Alpha) are Cineons/DPX. You may need LUTs for these files because of their non-linear colourspace.

Hope this helps.

Regards
Grant

DISCREET UK

[irte]

Hi
Thank you for your answer. I never heard before of a 32 bits per channel RGBA TGA format, and apparently all my tentatives to output such a format from MentalRay are failing. Are you sure you're not referring to the usual TGA32 format, which only has 8 bits for each RGBA channel?

I will now be a bit more precise, since I went a bit further with my investigation; I discovered I need, more than a usual RGBA format floating point image, a RGBE floating point image, which, not having maximum channel values clamped at 1.0 would allow me to store a higher dynamic range in it.
A newer question originates from this: is it possible for Combustion to work with float color values going past 1.0, or it just clamps them at 1.0 when computing color corrections and other operations? It would be great if it only clamped color values at the after compositing display stage.

More info for me please?
b

[dlm]

Hi,

That is what I mean by 32 bit TGAs. There are only 8 bits per channel. Sorry for the mistunderstanding.

When working in float mode in combustion, you are working between values of 0 to 1. You can adjust colour in a float image in decimal values which gives you a greater and almost unlimited amount of variation. Even if the values in the image are higher then 1 combustion will still keep this information. IT is just beyond the monitor and our own vision. To be able to see this information you would just darken the image and the detail in the whitest areas will still be there. No clamping occurs in float mode.

You may want to look at the SGI format which can accomadate float point.

Hope this helps.

Regards
Grant

DISCREET UK

[irte]

Hi
Thank you for your reply I am very happy to hear that values brighter than 1.0 will be maintained; now the problem is only about how to get them into Combustion
Well, let's see what I will be able to do with the SGI format.

Tnx again
b

[irte]

Just a little update:
I tried the SGI format, but the only thing I have been able to import in Combustion has been a 64 bit RGBA SGI image format. What I really need is a 16 or 32 bit RGBE image (in this case, the image channels have floating-point precision and their max value can exceed the max value of 1). When outputting a RGBE SGI image from MentalRay and importing it in Combustion, I got a weird-looking thing (clamped colors exceeding the allowed intensity range?).

So, still investigating...
b

[ryan3d]

What is a "floating point image" and why are you trying to do what youre doing?

[irte]

Floating point images are images with a much higher color resolution (so less prone to showing banding artifacts in example). The point is not having floating-point images in the comp, since in Combustion you can work in floating-point mode starting from regular 8 bit per channel images and get rid of the banding problem... the point is importing RGBE images, so images having floating point precision and color values with an intensity exceeding 1 (the so-called HDR images), in order to work with a much better reproduction of the intended luminance range.