Re: [Openexr-devel] (no subject)

[Richard Annema]

Hey all,

A document on how to read the files that the SplutterFish Max OpenEXR file
reader/writer produces is in the making - we're a bit busy with 3ds max 6
and v2 development of our main product at this time.
You should be able to decipher the file, though - the OpenEXR file writer
displays all the labels and datatypes used.

As for the UINT for RGBA - correct, it wasn't intended for it. The storage
is in a format that should be labeled as being 'packedRGBA'. Basically it
allows for 8bit/channel storage where OpenEXR does not normally support
8bit/channel (BYTE, or uhm.. QUART?) storage :)
To quote Steve when he added that.. "They made it extensible. I'm extending
it :)"
I'll ask him what the specifics of that storage format are, at least.

In my opinion, at least on the topic of all the extra channels, I think it
would be good to get some agreement across developers on what to label
channels and what attributes to set - I'm sure this will be talked about
more once we can get that document together.

For now, I bid you adieu with a peculiar Japanese device I forgot about
until I spotted it in our OpenEXR discussion logs, the CD-Poit!
http://www.watch.impress.co.jp/pc/docs/article/20011005/maxell.htm

Kindest regards,
Richard Annema
Director of Client Relations
SplutterFish, LLC

----- Original Message ----- 
From: "Drew Hess" <address@hidden>
To: "James McPhail" <address@hidden>
Cc: <address@hidden>
Sent: Monday, November 03, 2003 19:12
Subject: Re: [Openexr-devel] (no subject)


>
> Hi James,
>
> Yes, you can export any channel as UINT.
>
> How values are represented in a UINT channel is unspecified by the file
> format; it's up to the application to decide that.  You should ask the
> Splutterfish team about the method they use (or maybe they'll answer here,
> I think they read this mailing list).
>
> If you're using the 1.0.6 release of OpenEXR, the IlmImf library supports
> automatic datatype conversion for InputFile objects, but it doesn't try to
> be clever about scaling or normalization; it simply converts the data type
> (modulo representation issues like infinities, etc.).
>
> So, for example, if the file has a UINT channel and a sample value of 255,
> and you ask for a FLOAT channel in your frame buffer, readPixels will
> convert that to 255.0 float.  If you want that to be mapped to 1.0, you'll
> have to scale the float buffer after you get it back from readPixels ().
>
>
> -dwh-