[FFmpeg-devel] [PATCH] Add a gamma flag to exr loader to avoid banding

Gonzalo Garramuno ggarra13 at gmail.com
Tue Apr 29 01:21:03 CEST 2014


On 28/04/14 17:10, Michael Niedermayer wrote:
> On Mon, Apr 28, 2014 at 09:19:07PM +0200, Reimar Döffinger wrote:
>
>>> patience
I apologize for being so impatient.  Sorry.
>>> First it seems to still have a bug
>>> it fails fate
>>> --- ./tests/ref/fate/exr-slice-raw      2014-04-28 04:14:37.849375679 +0200
>>> +++ tests/data/fate/exr-slice-raw       2014-04-28 19:50:27.430558597 +0200
>>> @@ -1,2 +1,2 @@
>>> #tb 0: 1/25
>>> -0,          0,          0,        1,  3169800, 0x6a356d0d
>>> +0,          0,          0,        1,  3169800, 0x2cfefd25
I am stuck with this bug.  Not sure exactly what the numbers mean in the 
first place, except that it seems +0 and -0 are misplaced.

The fate test should fail with the exr_patch6.patch since gamma was set 
at a default of 2.2.  But that's not the case why it is failing.  I 
removed the powf function from the table lookup and that's not either.

IEEE math is not my strongest suit.  Can an IEEE guru take a look at the 
code, please (Reimar?)?  There is a HACK comment in the exr_flt2uint 
function which appears harmless, but not sure.

>> 3) It adds an option that people might rely on but which might be incompatible/require more hacks to support in a better solution. I would suggest marking the option as experimental/unsupported in the meaning we may remove it at any time with only a minor version bump.
> agree
>
> [...]
Agreed.


More information about the ffmpeg-devel mailing list