View Full Version : TFM vs TDeint?
TheBashar
31st August 2005, 02:45
I'm sure this is a stupid question, but I've spent 20 minutes reading through old threads and I still don't have a good feeling as to when TFM's internal deinterlace routines are better suited than using TDeint's. Can somebody help me out?
This is what I think, but I'm probably way off base. I think TFM's deinterlacer is not motion compensated. I think it is more sophisticated (than TDeint) at finding fields to match to produce as good of an image as possible. I think TDeint is less sophisticated about trying fields and uses a motion compensated method to interpolate a final frame. So, I think IF there is a good match to be found, TFM's internal is more likely to find it and the result will probably be higher quality than TDeint's interpolation. But if there just is no good match, TDeint will probably produce a better interpolation than TFM's best match.
Am I anywhere close?
Thanks!
Guest
31st August 2005, 02:50
TFM is a field matcher. It's for progressive content. If you have interlaced content there are no matching fields (except in static scenes, to be pedantic), so TDeint should be used. In this context, fields "match" if they are sampled from the same temporal moment. For interlaced video, all fields are sampled at different moments, so they cannot match.
You do know the difference between progressive and interlaced content, I assume.
TheBashar
31st August 2005, 03:36
It's fair to say I'm a bit confused. Evidentally what I should have referred to was "decombing" instead of deinterlacing. I have progressive content with combing artifacts. So I'm guessing the artifacts that I'm seeing are from the telecine process.
What I would like to understand is what are the implications of using TFM vs TDeint telecined progressive content?
neuro2, from your reply I take understand that TFM's field matching would not make sense on interlaced content while TDeint would be a reasonable approach. Am I right to assume that the converse would also be bad? (Using TDeint to remove combing from telecined progressive content)
Thanks for your help!
Guest
31st August 2005, 04:05
It's fair to say I'm a bit confused. Evidentally what I should have referred to was "decombing" instead of deinterlacing. I have progressive content with combing artifacts. So I'm guessing the artifacts that I'm seeing are from the telecine process. Instead of guessing, why don't you determine it definitively? If you post a link to a short unprocessed source sample, we can tell you.
Am I right to assume that the converse would also be bad? (Using TDeint to remove combing from telecined progressive content) Yes, it would not be as good as proper field matching, because proper field matching will restore the original progressive frames, while deinterlacing will interpolate. Also, if it is indeed 3:2 telecine, you'll get an objectionable strobing effect, because only 2 out of every 5 frames are combed.
TheBashar
31st August 2005, 11:44
Instead of guessing, why don't you determine it definitively? If you post a link to a short unprocessed source sample, we can tell you.
Thanks, I appreciate the offer. I'm sure it's mostly just telecide. I said "I guess" because in some of the videos I have there are bad cuts or something which result in combing. The vast majority match the telecine pattern.
Also, if it is indeed 3:2 telecine, you'll get an objectionable strobing effect, because only 2 out of every 5 frames are combed.
Thanks! I've been letting TDeint handle all my "combing artificats". I did not reallize that it would cause such an effect. I think I got confused because TFM's help file refers to "deinterlacing" throughout so I got it in my head that it and TDeint were doing the same thing. Thanks for setting me straight!
Cheers!
tritical
31st August 2005, 18:23
Thanks! I've been letting TDeint handle all my "combing artificats". I did not reallize that it would cause such an effect. I think I got confused because TFM's help file refers to "deinterlacing" throughout so I got it in my head that it and TDeint were doing the same thing. Thanks for setting me straight!
TFM's help file refers to deinterlacing frames that come out of field matching combed. The basic difference between tfm/tdeint is tfm first tries to match fields and then only deinterlaces frames that come out of field matching combed. TDeint, on the otherhand, deinterlaces everything by default... though it can be configured to first check frames for combing and only deinterlace combed frames. TDeint can also do very rudimentary field matching using the "tryweave" option, but it is nowhere near as good as tfm. Likewise, tfm can do deinterlacing, but it is nowhere near as good at deinterlacing a true interlaced stream as TDeint.
TheBashar
31st August 2005, 19:52
TDeint can also do very rudimentary field matching using the "tryweave" option, but it is nowhere near as good as tfm. Likewise, tfm can do deinterlacing, but it is nowhere near as good at deinterlacing a true interlaced stream as TDeint.
I'm just curious, why did you choose not to use the TDeint code for the deinterlacing part of TFM? (by deinterlacing, I'm referring to the decombing of frames that come out combed from the field matching process)
I, incorrectly, was using TDeint for everything. Since I've switched to TFM it looks like the results are much crisper for the majority of the clip, but there periodically is a combed frame at a cut, fade, or other edit place. I don't remember ever seeing a combed frame come out of TDeint.
Cheers!
TheBashar
2nd September 2005, 01:14
Since I've switched to TFM it looks like the results are much crisper for the majority of the clip, but there periodically is a combed frame at a cut, fade, or other edit place. I don't remember ever seeing a combed frame come out of TDeint.
I'm such a boob. Now that I understand the role of the deinterlacer in TFM (vs a full deinterlacer like TDeint) I can now appreciate how/why one would use TDeint in a post-processing only mode. I finally understand the use of TFM's clip2 paramter! Yay!
Cheers!
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.