[FFmpeg-devel] Enhancing/extending OpenEXR 'gamma' feature

[Kevin Wheatley]

On Fri, Aug 28, 2015 at 3:54 PM, Paul B Mahol <onemda at gmail.com> wrote:
> On 8/28/15, Kevin Wheatley <kevin.j.wheatley at gmail.com> wrote:
> Feel free to do it.
>
> Note however that those various curves really belong to avfilter once
> swscale supports float pixel format and exr decoder make use of it.

Absolutely, most of the tools I normally work with are full
float/double internally, I was going to start with the functions that
perform the non-linear encoding, but if I'm to consider the option
they will need to work in the swscale code it opens a few questions:

Should I provide functions going too and from linear? This would allow
them to be composed to perform cross conversion.
float or double precision? - if these functions will be composed
together is there a need for double precision?

Thanks

Kevin