Thanks for replying.
> I believe your problem could be rooted in bicubic transform.
> It makes n9 sense when source and target sizes are the same.
> Try to set linear.
I tried both SWS_FAST_BILINEAR and SWS_BILINEAR (as well as a few
others) and there is no change.
From what I understand of the sws_scale source code, when the source
size equals the destination size, the scaling operations are entirely
skipped, which makes sense for optimization.
It also makes sense since sws_scale's behavior is different if there's a
resize or not.
> Can you reproduce this with ffmpeg (the command line tool)?
> If yes, please provide command line together with complete,
> uncut console output.
Forgot to tell that in my first email:
I cannot reproduce this behavior, be it with ffmpeg, ffplay or VLC.
I am currently trying different options with ffmpeg, but as of now the
issue seems to be only in my project.
Also I've been thinking about the unusual size of the frames...
190 is not a factor of 8, while most standard resolutions are. And since
6 pixels are lost, there's only 184 pixels per line in a frame, which IS
a factor of 8.
I will be trying multiple resolutions, but I'm starting to think that
sws_scale has troubles handling width that can't be divided by 8.
Any thoughts on this?
Alex
P.S.: I'm not sure how this maillist is handled, so there's a chance
this message will show up on the wrong level in the web archive. Please
let me know if I should be doing something special.
_______________________________________________
Libav-user mailing list
[email protected]
http://ffmpeg.org/mailman/listinfo/libav-user
