Need some advice!

[xtremecanadian]

Hey guys,

It's been about 2 months since I've backed up a dvd. Since that time there has been some changes to dvd-rb and I would like to know if I'm up to par with everything. I feel as if I'm out of the loop with everything.

I like to stick with the basics to back up my movies. I use dvd-remake to remove buttons and some frames, dvd shrink to remove the audio, and dvd-rb v.66a with CCE SP v. 2.67, dgdecode, and decomb521. Personally, this method has always worked for me.

I know this question may sound stupid, but if I use these same methods with the latest version of rebuilder, will my results be the same? Should I get a newer version of CCE? or any of the .dll files?

I'm still reading to catch up on lost time.
Thanks guys

[jptheripper]

get the latest dgdecode

you should rarely use decomb (i never had in 100+ movies)

rebuilder removes audio just fine, i would skip shrink

definately get the newest rebuilder, lots of bug fixes


enjoy

[xtremecanadian]

So removing decomb completely is a good idea?

I just added it in because people recommended it.

[TheSeeker]

You can keep your setup page pointed to Decomb521 but just DONT USE the Deinterlace with Decomb option in the avs options menu. Get dgdecode version 1.10, get newest possible rebuilder, use rebuilder to remove audio, you dont need to use shrink at all. Other than that your backup strategy is the same as mine. Rip w/ decrypter, remove buttons and extras with Remake, run through rebuilder (but i use cce 2.70), and then I make an iso with ImgTool Classic, and I burn that with either Nero or decrypter.

[rhayman]

Originally posted by TheSeeker
You can keep your setup page pointed to Decomb521 but just DONT USE the Deinterlace with Decomb option in the avs options menu.
_______________

Im using DGdecode but i still have decomb521 set up and have used the deinterlace with decomb option in a couple of previous backups.

Why is this a bad thing?

Just out of interest when using the Convert 4:3 Letterbox to 16:9 option do you need to use deinterlace with decomb option or does Rebuilder take care of this automically.
I remember doing this manually once before and you could still see loads of lines on the tv. It was only an extra though so i did not look into it.

[Boulder]

Deinterlacing should always be the last choice when you encode for viewing on the TV. Interlacing is not a fault in the video.

If you see combing on the TV, there's something wrong with the encoded file, either an incorrect field order or improper encoder settings.

[jdobbs]

Originally posted by rhayman
Just out of interest when using the Convert 4:3 Letterbox to 16:9 option do you need to use deinterlace with decomb option or does Rebuilder take care of this automically.[/B] DVD-RB does the separatefields/resize/weave correctly for you when resizing interlaced material.

[Boulder]

That's the method you shouldn't use..especially when the difference is that big (720->352). You should use the bobbing method, preferably with a smart bobber.

[jdobbs]

?????? We're talking about resizing from 4:3 to 16:9, not Half-D1... the resolution stays the same.

[Boulder]

Well, applies to 4:3->16:9 conversions all the same;)

Does Half-D1 processing use bobbing?

[wmansir]

half-D1 is not a vertical resize, so de-interlacing isn't even needed. Of course it is possible to do, at the user's preference, but I don't see why bobbing would be preferred over other methods.

As for 3:4LB->16:9, as jdobbs said, if interlaced, it is currently resized by field to preserve the interlace. Of course, this resize isn't as sharp as a full frame progressive resize, but the user has the option of enabling deinterlacing. In that case the frame is deinterlaced and then treated as a progressive frame, which means it is not resized by field.

EDIT: D'oh, I meant not a vertical resize, but left out the "not" part. Thanks Trahald.

[Trahald]

Originally posted by wmansir
half-D1 is a vertical resize, so de-interlacing isn't even needed. Of course it is possible to do, at the user's preference, but I don't see why bobbing would be preferred over other methods. hmm? i thought 1/2 d1 is 352x480/576 ? (horizontal)

[Boulder]

Originally posted by wmansir
As for 3:4LB->16:9, as jdobbs said, if interlaced, it is currently resized by field to preserve the interlace. Of course, this resize isn't as sharp as a full frame progressive resize, but the user has the option of enabling deinterlacing. In that case the frame is deinterlaced and then treated as a progressive frame, which means it is not resized by field.

This is where Bob->Resize->Reinterlace should absolutely be used. You'll get vertically misaligned fields if you go the SeparateFields->Resize->Weave -route.

http://forum.doom9.org/showthread.php?s=&threadid=90950

I also think that the deinterlacer should be changed from FieldDeinterlace to TDeint or (Leak)KernelDeint since interpolation is used instead of blending. Those two deinterlacers keep much more detail since they are area-based instead of full-frame deinterlacers.

[wmansir]

I see what you are saying now, and yes that is a much better method for vertical interlaced resizing.

I don't really have an opinion on which deinterlacer to use because I don't generally use them. FieldDeinterlace seems tuned for speed, so I'm sure there are higher quality deinterlacers available.

[jdobbs]

[Modified] Hmmm... I thought about it and it has finally dawned on me that 4:3 -> 16:9 actually does vertical resizing as well.

So the recommendation is to do bob().lanczosresize().reinterlace?

I'll do some tests. Does this take a lot of time?

[Boulder]

It is ~50% slower as there's twice as many frames to process but it's still quite fast if you don't do any filtering. LeakKernelBob is apparently the fastest so I'd give that one a go first. Fortunately the situation shouldn't occur very often:)

[jdobbs]

reinterlace doesn't appear to be an inherent command. I can't add a requirement to have plug-ins... is there a way to do this completely with inherent AVISYNTH commands?

[Never mind] I got it. The example in the link showed it.

[jdobbs]

I just did a test of this using these lines on an interlaced segment so I could compare the output against a real source:

#------------------
# AVS File Created by DVD Rebuilder
# VOBID:01, CELLID:01
#------------------
LoadPlugin("C:\Program Files\DVD-Rip\DVD-RB\DGDecode.dll")
mpeg2source("D:\IMAGE4\D2VAVS\V02.D2V")
trim(0,27094)
a=separatefields().lanczosresize(720,240,0,30,720,180).weave()
b=bob().LanczosResize(720,480,0,60,720,360).separatefields().selectevery(4,0,3).weave()
stackhorizontal(a,b)
ConvertToYUY2(interlaced=true)
assumefps(4)
AudioDub(BlankClip())


I have to tell you that the "a" picture which uses separatefields/resize/weave is much clearer and more detailed. I think I'm going to stick with what my eyes tell me.

[wmansir]

I think your script has 'b' being a child of 'a'. Try:


#------------------
# AVS File Created by DVD Rebuilder
# VOBID:01, CELLID:01
#------------------
LoadPlugin("C:\Program Files\DVD-Rip\DVD-RB\DGDecode.dll")
mpeg2source("D:\IMAGE4\D2VAVS\V02.D2V")
source=trim(0,27094)
a=source.separatefields().lanczosresize(720,240,0,30,720,180).weave()
b=source.bob().LanczosResize(720,480,0,60,720,360).separatefields().selectevery(4,0,3).weave()
stackhorizontal(a,b)
ConvertToYUY2(interlaced=true)
assumefps(4)
AudioDub(BlankClip())

[Boulder]

Did you try LeakKernelBob, found in LeakKernelDeint.dll? A stupid bob is not the best choice since it treats the whole frame as interlaced. A smart bob treats only the parts that are found to be combed, the rest are treated as progressive. Give LeakKernelBob(order=1,sharp=true,threshold=7) a go and compare.

Also, Bob() is not lossless, on the other hand Bob(0,x) , where x=0.5 or 0.6 for example, is completely lossless when followed by SeparateFields().SelectEvery(4,0,3).Weave().

You should also use AssumexFF() in front of bobbing and right before reinterlacing, x depending on the field order which most likely is TFF. This is to ensure no field order changes occur. Avisynth assumes BFF by default for some weird reason.

ConverttoYUY2() should be used when you use a bobber, it can be placed right before reinterlacing. No need for the interlaced=true parameter as the frame's progressive.

[jdobbs]

Originally posted by wmansir
I think your script has 'b' being a child of 'a'. Try:


#------------------
# AVS File Created by DVD Rebuilder
# VOBID:01, CELLID:01
#------------------
LoadPlugin("C:\Program Files\DVD-Rip\DVD-RB\DGDecode.dll")
mpeg2source("D:\IMAGE4\D2VAVS\V02.D2V")
source=trim(0,27094)
a=source.separatefields().lanczosresize(720,240,0,30,720,180).weave()
b=source.bob().LanczosResize(720,480,0,60,720,360).separatefields().selectevery(4,0,3).weave()
stackhorizontal(a,b)
ConvertToYUY2(interlaced=true)
assumefps(4)
AudioDub(BlankClip()) I don't think so... I think "source" is assumed.

[jdobbs]

Did you try LeakKernelBob, found in LeakKernelDeint.dll? I'm not going to start adding a requirement to have certain filters in the avisynth plug-in directory to make DVD-RB work. I already made that mistake with decomb.dll and won't make it again. I'll be chasing snipe down rabbit holes for the rest of my life.

If someone wants to get really esoteric in his/her filtering -- they can use filter editor.

You should also use AssumexFF() in front of bobbing and right before reinterlacing, x depending on the field order which most likely is TFF. This is to ensure no field order changes occur. Avisynth assumes BFF by default for some weird reason. I put that line in, and it had no effect.

[jdobbs]

@boulder

Sorry... but it just looks better -- the changes you've suggested have no noticable effect.

I'd say you should give it a try yourself. The difference isn't incredibly dramatic, but the original method is clearly better on the sources I've tried.

Gotta get back to work on the Segment Editor I'm adding to Pro...

[wmansir]

Originally posted by jdobbs
I don't think so... I think "source" is assumed.

I see that you are right. I always thought 'last' was assigned to the last function, regardless of whether it was also given an explicit handle or not. Looks like I was wrong.

[jdobbs]

Don't worry about it. As can be proven with a simple scan of these forums... while I have never actually admitted to being wrong, I have come up considerably short of being right on many occasions... :D

[Boulder]

@jdobbs: Believe me, I have tried:) SeparateFields-Resizing-Weave gives me a softer image than Bob-Resize-Reinterlace. Recently I've had to fight a lot with fieldblended Star Trek TNG PAL captures, which made me experiment with the different options.

See the screenshots scharfis_brain posted (and he is one expert in this area) in that thread I posted the link to. Look at the Japanese subtitles - much sharper in the bobbed ones.

Actually Bob() uses SeparateFields() but also goes a little further:
IanB posted this one, it's what Avisynth does when you call Bob()

Global NewHeight=1234 # Whatever
Global NewWidth=1234 # Whatever else
SeparateFields()

Shift=(Height()/Float(NewHeight/2)-1.0)*0.25 # Field shift correction

Tf=SelectEven().LanczosResize(NewWidth, NewHeight/2, 0, -Shift, Width(), Height())
Bf=SelectOdd().LanczosResize(NewWidth, NewHeight/2, 0, Shift, Width(), Height())
Interleave(Tf, Bf)
Weave()

I tested that script of yours and from what I see, the sharpness seems to occur because that field shift is not being corrected. Thus, increased contrast near the edges, looks a bit like ringing to me.

You are, of course, the one who decides what should be used. I'm only tossing out some food for thought, being a Filter editor user myself;)

[jdobbs]

It's all cool... :cool: I'll do a few more tests.

[jdobbs]

See the screenshots scharfis_brain posted (and he is one expert in this area) in that thread I posted the link to. Look at the Japanese subtitles - much sharper in the bobbed ones. In his example there was definitely greater clarity to the "bob" image. But his example didn't match the 4:3 --> 16:9 conversion I was doing as it was aimed at emulating 1024i to 480i conversion, so I did my own comparison with the specific example of this specific type of conversion to be sure... and I just got a different result.

[Boulder]

I just found a very good test scene. See these screenshots, particularly the pattern on the man's shirt:

http://www.saunalahti.fi/sam08/org3.jpg
http://www.saunalahti.fi/sam08/separatefields.jpg
http://www.saunalahti.fi/sam08/bob.jpg

I used this script, interlaced PAL source:

LoadPlugin("C:\Utils\DVD Rebuilder\MPEG2Dec3dg.dll")
mpeg2source("D:\TEMP\DVD-RIP\REBUILDER\D2VAVS\V01.D2V",idct=7)
trim(1529,2968)
a=separatefields().lanczosresize(720,288,0,36,720,216).weave().ConvertToYUY2(interlaced=true).subtitle("sf")
b=assumetff().leakkernelbob(order=1).LanczosResize(720,576,0,72,720,432).assumetff().ConvertToYUY2().separatefields().selectevery(4,0,3).weave().subtitle("bob")
c=last.converttoyuy2(true)
interleave(a,b,c)
assumefps(4)
AudioDub(BlankClip())

A stupid bob doesn't cause nearly this big difference though, the images are nearly equal.

[jptheripper]

@jdobbs

segment editor?!??!?!?!

[Rockas]

When xtremecanadian asked some advice he certainly didn't thought what he was getting in to :)

Please be carefull with your posts... or you may never get a "give some advice" again :D

[jdobbs]

I can see a difference with the leakkernalbob of your example -- but I just ran your same script (adjusting sizing for NTSC) using the standard bob in AVISYNTH and it confirms my earlier tests... the "sf" frame looks clearly better. Like I said, I don't want to embed code into DVD-RB that forces a need for a third party filter... so for now I'm going to keep it as-is.

Thanks.

[Boulder]

Hey, no problem, you run the show here:)