[FFmpeg-devel] libavcodec/exr : add support for datawindow who exceed display window

Martin Vignali martin.vignali at gmail.com
Mon Apr 11 21:16:11 CEST 2016


2016-04-10 19:50 GMT+02:00 Martin Vignali <martin.vignali at gmail.com>:

>
>
> 2016-04-10 19:23 GMT+02:00 Paul B Mahol <onemda at gmail.com>:
>
>> On 4/10/16, Martin Vignali <martin.vignali at gmail.com> wrote:
>> > 2016-04-10 18:12 GMT+02:00 Paul B Mahol <onemda at gmail.com>:
>> >
>> >> On 4/10/16, Martin Vignali <martin.vignali at gmail.com> wrote:
>> >> > 2016-04-09 6:54 GMT+02:00 Paul B Mahol <onemda at gmail.com>:
>> >> >
>> >> >> On 4/9/16, Martin Vignali <martin.vignali at gmail.com> wrote:
>> >> >> > 2016-04-08 22:11 GMT+02:00 Paul B Mahol <onemda at gmail.com>:
>> >> >> >
>> >> >> >> On 4/7/16, Martin Vignali <martin.vignali at gmail.com> wrote:
>> >> >> >> > Hello,
>> >> >> >> >
>> >> >> >> > in attach, a patch with some modifications in size/data window
>> >> >> >> management,
>> >> >> >> > in order to manage
>> >> >> >> > negative datawindow/display window, and datawindow who can
>> exceed
>> >> >> >> > display
>> >> >> >> > window
>> >> >> >> >
>> >> >> >> > Size of the picture is now set from min and max value of the
>> >> display
>> >> >> >> window
>> >> >> >> > (before the patch, xmin, ymin of the display window are ignored
>> >> (and
>> >> >> >> assume
>> >> >> >> > to be 0))
>> >> >> >> >
>> >> >> >> > With the previous patch (fix channel detection), official
>> samples
>> >> >> >> > ./scanline/Blobbies.exr
>> >> >> >> > can now be correctly decode.
>> >> >> >> >
>> >> >> >> >
>> >> >> >> > DataWindow bigger than display window is a feature of the
>> >> >> >> > openExr.
>> >> >> >> > It can be use in some softwares, (for example to improve blur
>> >> >> >> > quality)
>> >> >> >> > Following the official documentation, for display, we need to
>> use
>> >> the
>> >> >> >> > display window only.
>> >> >> >> >
>> >> >> >> >
>> >> >> >> > In this patch, xmin, xmax, ymin, ymax, are now int32_t instead
>> of
>> >> >> >> uint32_t,
>> >> >> >> > because these values can be negative. (See officiel sample
>> >> >> >> ./Displaywindow
>> >> >> >> >
>> >> >> >>
>> >> >>
>> >>
>> http://download.savannah.nongnu.org/releases/openexr/openexr-images-1.7.0.tar.gz
>> >> >> >> > )
>> >> >> >> >
>> >> >> >> > Comments welcome
>> >> >> >>
>> >> >> >> Crashes with:
>> >> >> >>
>> >> >> >> openexr-images-1.7.0/DisplayWindow/t03.exr
>> >> >> >>
>> >> >> >> Strange, this sample didn't crash for me (Clang, OS 10.9). Decode
>> >> it's
>> >> >> > still not good after the patch but it's piz compression.
>> >> >> > Can you tell me, what configure line, you use ?
>> >> >>
>> >> >>
>> >> >> I still can't reproduce the crash for this sample.
>> >> > Is it the only sample who crash in openexr-images-1.7.0/DisplayWindow
>> >> > for
>> >> > you ?
>> >>
>> >> No, also t05.exr.
>> >>
>> >>
>> > Can you test with the new version of the patch in attach ?
>>
>> This one doesn't crash.
>>
>>
>> Thanks for testing.
>
> I remove the log line, in the patch in attach.
> I also add a test for wrong display window in decode_header func.
>
>
>
The last patch seems to have a trouble for some tile sample where
datawindow exceed display window (i think for scanline the last patch is
correct).
I work on it, and i will send a new patch soon.

Martin
Jokyo Images


More information about the ffmpeg-devel mailing list