[vdr] Re: Suggestion/Question

[Guido Fiala <gfiala@s.netic.de>]

On Friday,  2. November 2001 15:47, you wrote:
> On Fri, Nov 02, 2001 at 10:29:58AM +0100, Guido Fiala wrote:
> > Would'nt that make it possible to simply store some part of I-Frames in a
> > library (which is automatically build and increased by manually editing
> > recordings). Later recordings I-frames will then be checked for being in
> > library and only stored on disk if not...
>
> Maybe it will be enough only to compare some kind of CRC instead of the
> full I-Frame. Then you only have to search for a certain sequence of CRC's.
> I think in most (99.99999% :-) cases it will match with the right I-frame
> sequence.

Of course not the entire I-Frame :-), the advantage of the checksum (if it is 
_not_ CRC checksum but a MD5-fingerprint is it certaincy, but it' random and 
not compressable and probably slow to use.
On the other side, using some DCT-coefficients might make it not as "sharp" 
and allows to use IMHO faster code for the library (e.g. huffman?), however 
it makes the frame-scanner slower on the other side...

Finally - we can upload/download such libraries via internet and share it 
among us for later use!

To make the library not ever-growing, i would implement a "usage" counter, 
which automatically removes I-Frames from lib if they are no longer (or some 
time) not used at all.