[mythtv] PATCH: Commercial detection/skip using blank frames

Bruce Markey bjm at lvcm.com
Sat Feb 22 11:18:31 EST 2003

Matt Zimmerman wrote:
> On Sat, Feb 22, 2003 at 03:47:04AM -0800, Bruce Markey wrote:
>>Have you considered identifying these frames during recording rather than
>>playback? A cutlist file could be generated once so that jumps could be
>>immediate during playback without ever having to search.
> There are commercial detection algorithms which depend on being able to seek
> backward in time (my VCR appears to work this way), so at least the
> interface should not rule out this approach.

Right. I guess what I'm thinking is that it may be more
efficient to identify the blank frames on a single pass
(during record). After building a list of all the blank
frames, it may be easier to accurately identify commercials.
One possibility is to sample the frames while recording
then analyze and build a cutlist after the recording has
ended. Even storing the list of blank frames without a
cutlist would be faster during playback since seeking
wouldn't be necessary.

The current method is slow for remote frontends because
all the data has to pulled across the network before it
can be can be sampled. This can take 15-30 sec for a single
commercial skip.

--  bjm

More information about the mythtv-dev mailing list