On 2009-01-21, David Schleef <[email protected]> wrote:
> On Tue, Jan 20, 2009 at 11:42:31PM +0000, [email protected] wrote:
>> From: David Flynn <[email protected]>
>>
>> Also removed the have_sequence_header flag, since it is identical
>> to sequence_header_buffer != NULL;
>
> I would prefer to use have_sequence_header, and not keep the sequence
> header buffer around. The latter is only used to do a bit-exact
> match to subsequent sequence headers, which doesn't expand to everything
> it needs to do. For example, it doesn't handle comparing everything
> except interlaced_coding.
The only use i see for sequence_header_buffer is for confidence scoring:
- stash the first seqhdr you see
- have a saturating counter that:
- saturates at say 3.
- increments for each subsequent seqhdr that matches
- decrement for each failed match
- if the counter ever underflows, assume there is a missed EOS, etc.,
create a new instance and destroy the current one.
Other factors can be used to influence the confidence, eg step changes
in picture number range. (not impulses)
However, it doesn't help much for deciding if a new sequence is pretty
much identical to the old one, something that i was wanting to use.
Status: Unknown -- shall we implement the above?
..david
------------------------------------------------------------------------------
This SF.net email is sponsored by:
SourcForge Community
SourceForge wants to tell your story.
http://p.sf.net/sfu/sf-spreadtheword
_______________________________________________
Schrodinger-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/schrodinger-devel
