Undot() by default?

[Alxemi]

Hi all
I´ve seen that now all the Gordian Knot AVS scripts have the Undot() function by default.. what do you think about it? are there always dots to be removed?
And more generally, do you use any other functions by default? (Lumafilter, noise removing, etc)

[NightMare]

well undot helps alot in alot of movies i loved it when i encoded the godfather

but i dont think it should be on by defult there should be an option for it

by the way wrong forum post in gk :D

[Alxemi]

by the way wrong forum post in gk
My question is not about gk using, is about avisynth using... gk is only a excuse to start the question :D
What really means is what the people thinks about using a filter (any filter) by default, like Undot() or Lumafilter() or whatever()

[Manao]

If I had to choose a default filter, it would be Convolution3D(preset="movieHQ"). I tried once UnDot(), and didn't like the result I got. C3D with this setting is very subtle, yet very effective.

[Si]

I tried once UnDot(), and didn't like the result I got.

I'd be very interested in seeing a clip that UnDot could do any harm to ;)

It really goes out of its way to only do good :)

regards
Simon

[qwerpoi]

I agree with siwalters - I've never seen a clip where Undot() had any undesired effects (i.e., it just got rid of dots). Maybe it would be bad if the image is supposed to look grainy, I'm not sure. If you want to be sure of what the filter is doing, use the Avisynth Subtract() function, i.e.,

a = Avisource("C:\something.avi")
b = a.Undot()
Subtract(a,b)

Every time I used Undot(), I did a test like this, and every time I only see dots being removed. As far as making it a default, I personally do it with a lot of my scripts, but I don't think many people will agree it should be defaulted on all scripts. A filter is a filter, and there will be some off case where a filter will negatively affect your image, even if it is 1 in a million. As a rule you should always test to see if the filter is doing what you want it to do before you decide to use it.

[ADLANCAS]

@qwerpoi
nice trick!

Do you have an idea how to use with interlaced sources?

thanks,

Alexandre

[Boulder]

This is how to treat interlaced material:

SeparateFields()
#all spatial filters here#
Weave()

[leadman584]

UnDot() by default, bout time. I Have used this wonderful tool since I first learned of it. The cost to encoding speed is <2%, increase in compressibility is >10%, I've seen improvements >27% in some of my own encodes. CG Animation is the only source I have ever seen that doesn't necessarily need it, but it don't hurt either. Again it is worth it just for the compressibility gain.

[WildCelt]

Originally posted by Boulder
This is how to treat interlaced material:

SeparateFields()
#all spatial filters here#
Weave()

Does this also include telecined material?

[Boulder]

Originally posted by WildCelt
Does this also include telecined material?

I can't say anything for sure since I only deal with PAL material, but I think you should IVTC before doing any filtering.

[WildCelt]

I am going right to DVD, so I typically do not IVTC.

[Boulder]

Like I said, no knowledge in NTSC stuff :D

I thought that you had to IVTC to get a 23.976fps video and then do a pulldown..

[crusty]

I tried once UnDot(), and didn't like the result I got.
Could you tell us some more? I always use Undot(). Twice, actually.
Once before and once after resizing. Not by some rocket science or so, I hypothesizing that after resizing new dots will be created by the resizing process, that can be gotten rid of by another Undot().

You can use Undot() as many times as you want in an avisynth script.
This is how to treat interlaced material:
SeparateFields()
#all spatial filters here#
Weave()


Be sure to use only SPATIAL filters. I tried this once with Convolution3D and it *&^$%#-up my encode. Even when you tell C3D to only use spatial filtering it won't work.

[Boulder]

Originally posted by crusty

Be sure to use only SPATIAL filters. I tried this once with Convolution3D and it *&^$%#-up my encode. Even when you tell C3D to only use spatial filtering it won't work.

Yep, that's true.

Slightly OT, but all filters should work with this:

SeparateFields()
even=SelectEven().Convolution3d or whatever here
odd=SelectOdd().Convolution3d or whatever here
Interleave(even,odd)
Weave()

Just make sure you use the same settings for both even and odd frames;)

[mikeson]

Originally posted by crusty
Be sure to use only SPATIAL filters. I tried this once with Convolution3D and it *&^$%#-up my encode. Even when you tell C3D to only use spatial filtering it won't work.

Can you please tell me why? I often use Undot together with Convolution3D and I haven't seen any problem yet. Maybe some screenshots would help...

[kempfand]

I tried this once with Convolution3D and it *&^$%#-up my encode. Even when you tell C3D to only use spatial filtering it won't work. and
SeparateFields()
even=SelectEven().Convolution3d or whatever here
odd=SelectOdd().Convolution3d or whatever here
Interleave(even,odd)
Weave()This didn't work for me either (DV interlaced PAL). I then used Fold/Unfold from stickyboy (http://forum.doom9.org/showthread.php?threadid=60530&highlight=stickyboy) and this made C3D work like a charm.

Kind regards,
Andreas

[mikeson]

And what about progressive sources? I mainly encode PAL progressive and using Undot together with Convolution3D never gives me bad encode (at least I've never seen any artifacts). Any expreinces with this combination?

[Manao]

@crusty : I was using convolution3D, but I don't think they can't be used together

@all : When I said I was displeased by the result, I meant that I prefered the result with C3D alone. It may have been particular to the movie I was encoding.

[WildCelt]

Originally posted by Boulder
Like I said, no knowledge in NTSC stuff :D

I thought that you had to IVTC to get a 23.976fps video and then do a pulldown..

I dunno, that is news to me. I understood that if you were going straight to NTSC DVD that you shouldn't IVTC or deinterlace.

Can anyone shed some light on this?

[Erik_Osterholm]

I dunno, that is news to me. I understood that if you were going straight to NTSC DVD that you shouldn't IVTC or deinterlace.

I can't think of a time when it would hurt to do so (can half-D1 use pulldown?)

DVDs can be progressive source, but they'll need pulldown when playback, either by using pulldown.exe or the appropriate settings in your encoder. Encoding a progressive stream means better quality at equivalent bitrates when compared to interlaced or telecined material.

[WildCelt]

Ah good point. I never considered the bitrate angle! Thanks

[Erik_Osterholm]

I should mention that this requires a proper IVTC to work. If the source can't be IVTC'd correctly, you will obviously experience less than fluid motion. Search for threads on IVTC'ing Star Trek for more information. For absolutely highest quality, you probably need to leave these streams interlaced if you plan on viewing them through a standalone. But most shows/DVDs I've run across are IVTCable without much issue...I have more trouble on my own captures than DVD rips.