View Full Version : usage of FieldDeinterlace(full=false) on progressive source
len0x
4th November 2003, 19:02
I was wondering if FieldDeinterlace can be tuned to be used with progressive source so that it doesn't do much damage? full=false option should handle this case but I wonder what should be reasonable tresholds for DVD sources.
P.S. the question arises from an attempt to do automatic deinterlacing which means that progressive sources have to be handled properly as well in case deinterlacer is always applied...
P.P.S. Does anyone know a filter which can return probability of a given frame to be interlaced ? (or for sure, but this is too much to ask I think)
Si
4th November 2003, 21:46
I think that KernelDeInt would be better for this task. I use it as a post-processor on my primarly progressive material.
It has a variable threshold with a motion map and a one-way mode so that it doesn't make a mess of scene-changes.
Check out the neuron2's forum (http://neuron2.net/ipw-web/bulletin/bb/index.php)
Having said that, we are all looking for an AI program that can detect combing in mouth areas only :)
regards
Simon
PPS Lots of filters can return a probability - but all of them will foul up at least once during a clip ;)
len0x
5th November 2003, 16:07
Thanks, I'll give KernelDeInt a try.
Actually I'm surprised nobody had an attempt to do the following:
As there are filters which have combing detection inside already it should be possible to build a filter which scans source for high motion scenes and examines portions of 5 frames in a row to see whether they all are interlaced (pure interlaced source) or just two of them (telecined source). So you can make decision on which filter to use right after running a detector first right in the AVS script...
*EDIT* look what I found on Neuron's forum: http://neuron2.net/ipw-web/bulletin/bb/viewtopic.php?t=229
*EDIT2* moreover, this thread is partially describing what I was looking for :http://neuron2.net/ipw-web/bulletin/bb/viewtopic.php?t=207 nice job sh0dan. I'm not sure if I can use an arbitrary frame number to the paremeter of IsCombed() though...
Si
5th November 2003, 21:16
Which bit of
Lots of filters can return a probability - but all of them will foul up at least once during a clip
did you not understand :)
There are many ways to acheive a partial success of what you want - no method currrently known around here can fully handle "real world" avi/mpegs.
Lots of people have spent a lot of time searching for it.
Let us know if you find it :)
regards
Simon
len0x
6th November 2003, 11:12
Originally posted by siwalters
Let us know if you find it :)
Let me rephrase what I'm looking for: a method for automatic DVD-2-AVI(i.e. MPEG2-2-MPEG4) conversions which will work say 98% of the time. I know that it's possible to understand lots of things about the source just by looking at it. What'm trying to do is to replace a human eye in most of the cases (i.e. to find out either source is progressive or telecined. and don't tell me you can't to that manually in most of the cases). I'm not trying to achieve perfect decombing on any source you can imagine. I hope this makes more sense now.
vBulletin® v3.8.5, Copyright ©2000-2012, Jelsoft Enterprises Ltd.