smarter frame selection

[GoogleCodeExporter]

It would be nice to implement this algorithm:

http://notbrainsurgery.livejournal.com/29773.html

Original issue reported on code.google.com by [email protected] on 26 Jan 2009 at 5:16

[GoogleCodeExporter]

Looks interesting.

I can't promise any results soon, but I'll see if I can find some free time in 
the
near future.

Original comment by dirk.vdb on 26 Jan 2009 at 9:06

• Changed state: Accepted

[GoogleCodeExporter]

The code in svn contains an implementation of the algorithm. You can activate 
it with
the -p flag.

I have only one file here to test on, so I'm not convinced yet. But you can try 
it
out. If you get interesting results, please let me know.

Regards,
Dirk

Original comment by dirk.vdb on 5 Feb 2009 at 1:17

• Changed state: Started
• Added labels: Type-Enhancement
• Removed labels: Type-Defect

[GoogleCodeExporter]

I've tried it. It's good, especially with small videos. But when the video gets 
a
little bigger it seems to continue to sometime select black/dark frames. Here 
is a
file which may help: http://yozora-irc.net/~ubitux/pub-dove-onslaught.flv

While -t 15 gives a pretty good preview, -p seems to select more something like 
-t 10.

Also, you can try this bigger file:
http://movies.apple.com/movies/fox/avatar/avatar2009aug0820a-tsr_1080p.mov

Finally, it's generally better, and I'm pretty satisfied about that feature. 
Thanks!

Original comment by [email protected] on 2 Sep 2009 at 9:20

[GoogleCodeExporter]

I suppose it will give better results if -p switch will select starting frame 
depending on -t switch. In version 2.0.6, it builds histogram from the 
beginning of the file as I can see.

Original comment by [email protected] on 7 Feb 2011 at 9:14

[GoogleCodeExporter]

The -p option seems to give great results most of the time. The rest of the 
time it produces a black frame. It should be modified to not output a frame 
that is close to total black or total white.

Original comment by [email protected] on 3 Apr 2011 at 4:51