[FFmpeg-devel] Photo-sensitive epilepsy / 'Harding' test

compn tempn at twmi.rr.com
Sat Oct 5 03:30:29 CEST 2013

On Fri, 4 Oct 2013 23:18:38 +0100, John Warburton wrote:
>I notice that the Ofcom (UK broadcast regulator) guidelines on
>photo-sensitive epilepsy have been recently codified, and the description
>is FAR more precise than has been available before:
>Please look at Annex A in this document:

neat. thanks for posting.

>The Digital Production Partnership, which sets delivery standards for all
>major UK broadcasters, states in its latest draft that tape delivery
>requires a "Harding" analyser report to be submitted. However, for file
>delivery, an "approved device" must be used to test the footage. At the
>present stage of the published draft, there isn't a device specified. Is it
>possible that ffmpeg could include a libavfilter filter that checks for
>everything mentioned in Annex 1 of the document referred above? And, given
>the enormous trust (whether documented or undocumented) of ffmpeg in UK
>broadcasters, could ffmpeg achieve the status of an "approved device"?

yes, i think they would approve of such thing as long as it met the
goals and works with whatever program/deck (avid, blackmagic, etc)
and codec/format/colorspace broadcasters use.

>I'm happy to learn more C — I get a little lost with pointers at this early
>stage — to help program this. The documentation for libavfilter seems to
>offer me a way to perform analysis in the spacial and temporal dimensions

libavfilter can do it. just a matter of code.

>on video, both of which this analysis requires: how should I begin
>development in an open way, to show other developers my code without
>committing to the trunk?

start a free clone ffmpeg repo on github website. then just paste your
github url into a mail here.

i'm not epileptic, but the bright flashes in video annoy the shit out
of me. i'd be very happy to get rid of them. most sports programs in
the usa use extensive flashes to grab attention. those broadcasts would
be useful for samples as well.

heres an irritating sample:

and the ffmpeg ticket:
which has the infamous pokemon flashing video.

my idea was to have the filter duplicate the last frame for the
duration of the flash, to keep audio sync. obviously this isnt part of
what you need (the harding test), but i'd like to have that feature. if
the two frames differed too much, say a black frame to a white frame,
they could be morphed slowly to minimize quick bright flashes.


More information about the ffmpeg-devel mailing list