Re: [Openexr-devel] 16-bit unsigned integer

[Drew Hess]

Hi Robin,

On Tue, 15 Jul 2003, Robin Rowe wrote:

> I'm not sure I understand the implication. Is it your advice to accept the
> data loss from squeezing an unsigned short into the range of 0 to 1.0 in a
> half?


With half, you get almost 14 bits of resolution from 0 to 1.0, and you get
the ability to represent values greater than 1.  That's a good tradeoff
for storing HDR color images, and that's the purpose for which OpenEXR was
designed.


> I understand your reservation about adding more pixel types to a design, and
> if I was asking for a significant difference such as 8-bit operation I could
> better understand your reasoning. However, with 16-bit unsigned and 16-bit
> half why would there be any overhead supporting both? Are you doing
> something to 16-bit half in the act of reading it? Isn't it just data?


It's not a question of overhead in the file format.  We simply didn't need 
that data type when we designed OpenEXR.  

If you want to store 16-bit unsigned integer data, perhaps you should use
PNG or TIFF.  OpenEXR doesn't claim to be an appropriate format for all
data types.


-dwh-