Just for my understanding... with qtblend we would get:* luma wipes without the
luma bleeding that composite has,* affine functionality (with the
aforementioned luma wipes on top)* performance optimization with respect to
affine and composite, especially when B frames have no transparency
Did I got this somehow correct?
Best regards, Harald
-------- Ursprüngliche Nachricht --------
Von: Jean-Baptiste Mardelle <j...@kdenlive.org>
Datum: 15.07.2016 09:35 (GMT+01:00)
An: mlt-devel@lists.sourceforge.net
Betreff: Re: [Mlt-devel] Compositing transition question
On Friday, July 15, 2016 4:27:04 AM CEST, Brian Matherly wrote:
> JB,
>
> I think this optimization could be made in frei0r.cairoblend.
>
> Could you compare the performance of frei0r.cairoblend with
> transition_qtblend when the top frame has transparency?
>
> Seems like the transition_qtblend would be worth keeping
> regardless of whether its performance is better.
I did make some further optimisations to my qtblend transition, and
performance seems quite similar to cairoblend when doing a composite
operation.
qtblend can easily be modified to also use opacity and affine like
transformations by applying a QTransform to the QPainter.
I did spend some more time looking at cairoblend, composite and affine
transitions though, and here are my quick thoughts:
* composite transition: like Dan said, using the Kdenlive's monitor zoom I
can clearly see the luma bleed
* affine transition: when used as a filter, it does not correctly handle
alpha channel (see picture in http://j-b-m.ch/affine-bug.png ). I tried to
fix it but non trivial to fix for all cases
* frei0r.cairoblend: if we want to optimize it as I did for qtblend, it
would require modifications to transition_frei0r.c, since this helper
always requests RGBA images for the A and B frames. Making it retrieve only
B frame if it has no transparency does not seem easy since as Dan noted
some frei0r transition might want to use A frame even when B frame is fully
opaque, so it would require some kind rules depending on which
transition/parameters.
So in conclusion, despite the fact that I agree it's better to fix the
problems rather than adding a new solution, in this case - thinking about
Kdenlive - I think the best option would be to enhance my current
implementation of qtblend to allow affine like transform and use it as a
replacement for composite/affine/cairoblend in Kdenlive.
Regards
jb
------------------------------------------------------------------------------
What NetFlow Analyzer can do for you? Monitors network bandwidth and traffic
patterns at an interface-level. Reveals which users, apps, and protocols are
consuming the most bandwidth. Provides multi-vendor support for NetFlow,
J-Flow, sFlow and other flows. Make informed decisions using capacity planning
reports.http://sdm.link/zohodev2dev
_______________________________________________
Mlt-devel mailing list
Mlt-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mlt-devel
------------------------------------------------------------------------------
What NetFlow Analyzer can do for you? Monitors network bandwidth and traffic
patterns at an interface-level. Reveals which users, apps, and protocols are
consuming the most bandwidth. Provides multi-vendor support for NetFlow,
J-Flow, sFlow and other flows. Make informed decisions using capacity planning
reports.http://sdm.link/zohodev2dev
_______________________________________________
Mlt-devel mailing list
Mlt-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mlt-devel
