Laserdisc to DVD Transfer, encoding resolution?

[Piper]

First, I'm brand new to Laserdisc's, please bear with me. :-)

In the last few days I've obtained my first Laserdisc & Player with the intent of tranfering my favourite movie onto DVD. Having no previous experience with Laserdisc, I was curious what the quality would be like compared with my old vhs copy of the same movie.

The player (borrowed) came with a copy of WaterWorld (horrible horrible) with it. I played it back and was impressed with the visual quality of the picture. I was less impressed with the visual quality of my movie (to remain nameless) when I played it using the same setup. The transfer to Laserdisc was poorly done and there's a lot of noise in the picture. Nevertheless, I proceeded to capture it and throw my AVISynth scripts at it which cleaned it up significantly.

Here's where my inexperience with Laserdisc and DVD encoding comes into play, as up until now, I've been dealing mostly with VCD and DivX encodes.

I've read that pre 1990, Laserdisc (NTSC) resolution is 528x480 and 544x480 post 1990. The disc was pressed in 1989. Valid DVD resolutions I've read are, 1/2 D1 (352x480), 3/4 D1 (544x480) and Full D1 (720x480). Ok, we have a match at 544x480. What I'm finding however is that any resolution beyond 352x480 and I'm getting artifacts into the picture, as though I'm stretching the image beyond it's source resolution. Is the 528 or 544x480 Laserdisc resolution locked? Could it be just the result of a poor transfer (orignally)? If it looks good at 352x480, would you recommend I go with it rather than something higher? It should go without saying that I'm attempting to achieve the highest quality transfer I'm able.

If anyone else can provide some of their experiences with Laserdisc to DVD transfer, please let me know.

Thanks in advance

[DIggedy]

Valid DVD resolutions I've read are, 1/2 D1 (352x480), 3/4 D1 (544x480) and Full D1 (720x480)

Where did you read this? I've not heard of 544x480 as a dvd compliant resolution before.

Is the LD you're trying to transfer in widescreen? If so what ratio? The reason I'm asking this is if it is widescreen you should be able to use 352x480, stretch it into an 16:9 anamorphic image and still maintain most of the LDs original resolution.

[Piper]

It was in the vcdhelp forums (http://www.vcdhelp.com/forum/userguides/94382.php)

My mistake, after re-reading the article, it implies 3/4 D1 as a valid Laserdisc resolution, not DVD. So, it appears I'm limited to 352x480 or 720x480. If someone else can shed more light on this, please do so.

The LD's aspect ratio is 4:3.

Thanks.

[auenf]

even tho the source is 5xx x 480, its played back stretched to 4:3 (640x480).

352x480 is also played back stretched at 4:3, and 720(704)x480 is actually shrunk a little (or stretched, depending on the player).

of course if you tick 16:9 instead, it stretches it to 848 instead ;)

if you encode to 352, your going to lose quality, if you encode to 720(704) you arent going to gain quality, but you wont lose very much (if any) quality.

use 352 if you want to fit a lot of footage on the one DVD-R, otherwise use 720(704).

Enf...

[Piper]

I've been doing a lot of experimentation lately and I've found with some agressive Convolution3D filtering thrown at it, I've been able to get very good results with TMPGEnc at 720x480. I'll stay at that resolution now that I've been able to clean it up. Sometimes reducing the resolution is in itself a form of noise reduction.

Now to pick my encoder. I like what I'm seeing with the latest version of TMPGEnc, but I've heard a lot about how good the CCE and MainConcept codecs are. Can anyone point out a current MPEG-2 comparison site?

Thanks

[DIggedy]

Convolution3D

What is this?

[auenf]

Originally posted by Piper
I've been doing a lot of experimentation lately and I've found with some agressive Convolution3D filtering thrown at it, I've been able to get very good results with TMPGEnc at 720x480. I'll stay at that resolution now that I've been able to clean it up. Sometimes reducing the resolution is in itself a form of noise reduction.

Now to pick my encoder. I like what I'm seeing with the latest version of TMPGEnc, but I've heard a lot about how good the CCE and MainConcept codecs are. Can anyone point out a current MPEG-2 comparison site?

Thanks

doom9 did a comparison of procoder, cce and tmpgenc in a thread here (either this forum, or the CCE forum), 8mbit iirc.

do a search for it, otherwise theres a large compressor comparison at the same site where you get bitrate viewer.

Enf...

[Piper]

Originally posted by DIggedy


What is this?

In it's most basic form, a noise reduction filter. Check out the AVISYnth forums for more info.

[Jeff D]

My question is basically along the same lines, I thought I'd ask here instead of starting a new thread.

But, before I get started, auenf stated 16:9 has a virtual horizontal resoultion of 848, is this correct? I learned it as 853, only a difference of 5 but, could make a difference on a 2.35:1 movie. =)


Here are my questions.... I'm archiving some LDs to DVD and here's what I've got:
Captured video in virtualdub with cropped size of 720x276 (2.35:1 source) 30fps, audio 44k 16bit stereo. Capture done there were 80 dropped frames over 75 minutes, I hope that will be fine.

I've got an avisynth script that will take the 720x276 image and do the inverse telecine, and remove dup frames to get a 24 fps source, resize for anamorphic and add borders.

I'm serving that avs file up to CCE and trying to get the settings correct to give me a 24 frame mpeg video file.

The 3:2 pulldown detection is confusing me. Avisynth is serving up a 24fps source. Do I need to check this option? I do check the progressive checkbox in the video setting dialog.

Scenarist complained about the 23.97 + 1001 frame rate. I've been zlooking around an it looks like I'll want to run pulldown on the video stream file.

Can anyone give me advice on what I "should" be doing? Experimenting is taking a long time. And, it seems most of the encoding talk going on around here is for SVCD and VCDs, not really related to DVD which can handle a 23.97 fps video stream.

Thanks!

[slk001]

1) Capture at the MAXIMUM resolution and MAXIMUM bitrate that your capture card can do reliably (I can cap at 9.5Mbps at 704x480 using a GOP="IPPP" with no frame drops).

2) Perform a single pass IVTC encode (I use AVISYNTH, with the filter DECOMB, a GOP="IPPP", CBR at 9.5Mbps) to tranform your video to 23.976fps FILM.

3) Now, use THIS video to apply any filters, resizing, etc, then encode to VBR using encoder of choice (I use a 3 pass encode).

4) Apply a 2:3 pulldown, setting each frame as progressive, drop_frame TRUE, etc, etc.

5) Import and author as usual.

[Jeff D]

slk001, thanks. I've got followups:

1) Capture at the MAXIMUM resolution and MAXIMUM bitrate that your capture card can do reliably (I can cap at 9.5Mbps at 704x480 using a GOP="IPPP" with no frame drops).

I guess I forgot to mention I'm using HuffYUV for the capture. I figure that's better than any encoded format.


2) Perform a single pass IVTC encode (I use AVISYNTH, with the filter DECOMB, a GOP="IPPP", CBR at 9.5Mbps) to tranform your video to 23.976fps FILM.


My avisynth script to serve up the foloowing:
Trims the crap out
telecine()
decimate(5)

tx = 720
ty = 368

topborder = (((480-ty)/2)/16)*16 # top letterboxing rect, rounded t 16 pixels
BicubicResize( tx, ty )
AddBorders( 0,topborder,0,480 - (topborder + ty) ) # add black letterbox bars to 704 x 480 frame size

AssumeFPS( 23.976 )


3) Now, use THIS video to apply any filters, resizing, etc, then encode to VBR using encoder of choice (I use a 3 pass encode).

The script above goes off to CCE and VBR 3 pass. I tweeked the GOP structure to be
IBPBPBPBPB
with these features:
Close GOP
DVD Compliant

Video settings have:
Progressive Frame set to True
Size 16:9
(I thought there was a frame rate settings, but I don't remember for sure, if there was it was set to 23.976)


4) Apply a 2:3 pulldown, setting each frame as progressive, drop_frame TRUE, etc, etc.


What should I do this with pulldown.exe? Should I also NOT flag the prograssive frame and 16:9 in CCE?

When you say apply a 3:2 pulldown, you mean set the flags, correct? Doing a 3:2 pulldown will create a 29.97fps stream, I don't want that.

What else is "etc, etc"? There were a lot of options in pulldown that looked useful, but redundant to a stream that should be pretty close to "correctly" flagged.

[MickeyNBK]

What should I do this with pulldown.exe?

Run Pulldown yourfile.mpv youfile.m2v
Should I also NOT flag the prograssive frame and 16:9 in CCE?

You should still flag the progressive frame and set to 16:9



When you say apply a 3:2 pulldown, you mean set the flags, correct? Doing a 3:2 pulldown will create a 29.97fps stream, I don't want that.

Please explain why you do not want to do this. If you are making an NTSC DVD you have to use pulldown to make it 29.97fps. This will not change the size or time of your video. Everything will be as it was only this time Scenarist will allow you to import it. Your stream is still effectivly 23fps only now it's telling Scenarist that it's 29.97.

Trust me everything will be fine.

[slk001]

You have added borders to bring your video to 720x480. This is 4:3 resolution, NOT 16:9. If possible, capture at 853x480 resolution, then resize to 720x480, then set the 16:9 flag (this would be an anamorphic (?) video).

Anywhere you can set the PROGRESSIVE FRAME flag is fine. I use the pulldown.exe GUI, so I set:
FRAMERATE 29.97 (this only sets the display framerate, not the encoded rate).
PROG_FRAMES P
DROP_FRAME TRUE
ASPECT_RATIO 4:3

Of course, check the docs for proper syntax.

No need to close the GOPs. This is only needed for multiangle DVDs.

Your GOP N/M structure is not very efficient (only 1 "B" frame). Most commercial DVDs have a 3/4 structure, which will give you "IBBPBBPBBPBB". This is what I use.

[Jeff D]

Thanks guys, I'm rethinking the B frames, something about two consecutive B frames are not useful since a B frame can't be used to build off a preceding B frame (maybe this is wrong).

Last night I tried again in CCE with some new settings. I messed with the GOP, but didn't add the extra B frame in.

In CCE, I set the following:
3:2 pulldown detection (not sure how good this is, AVISynth was suppose to serve up a progressive stream)
Letterbox hint
16:9 Aspect


Under the quality settings I used:
Natural Image
slider more toward Difficult than easy
Progressive was checked, although redundant, the 3:2 pulldown detection sets this.

3 Pass VBR 2k min, 6k avg, 8k max
different files on each pass

The results looked pretty good, before passing it through pulldown.

(I have a better understanding of what pulldown.exe does, only sets flags, and I understand why drop_frame true is important) There is still a question of progressive sequences.

pulldown.exe should be run to do the following.
- Set 3:2 flags
- Set progressive frames
- Set 29.97 frame rate
- Set 16:9 aspect ratio

(note I can do some of this elsewhere, or here.) I'll repeat, I don't get how progressive sequence is used, and if it should be set.



You have added borders to bring your video to 720x480. This is 4:3 resolution, NOT 16:9. If possible, capture at 853x480 resolution, then resize to 720x480, then set the 16:9 flag (this would be an anamorphic (?) video).

I'll repeate the steps and maybe that will make sense:
- capture 2.35:1 source at 720x276. Actuall image, just cropped to the letterboxed image.
- use avisynth to resize, and add borders to the captured image. The image will be streched from 720x368 (this will keep the correct aspect when 4:3 is stretched (anamorphic stretch) to 16:9. Anamorphic strech does nothing to vertical size, that's why it must be done before encode. This is the same method used for creating DVD's, no?

I think what you are describing is what I'm doing, but in a different order. I can't capture at that resolution you suggested.

[Jeff D]

I do have a couple more question...

I've been concerned with the capture dropped frame (approx, 1/minute) I'm using the telecide and decimate filters in avisynth to reconstruct the 24fps source.

Is it a good idea to try the following
Telecide
Convolution3d(preset="movieHQ")
Decimate

My hope would be the Convolution3d would be used to clean up noise across the frames before the decimate.

Is this dumb?

The results are pretty good with my single pass VBR. I'd just like to get a cleaner image for display on the big screen. -)

Thanks again! I hope others find this helpful too!

[MickeyNBK]

Convolution3d(preset="movieHQ")

How does the output look so far? My guess (without seeing it) would be to use:
Convolution3d(preset="movieLQ")
HQ is for High Quality DVD video, since this is an analog capture from LD my guess would be to use LQ(low quality DVD). Try encoding a couple of minutes worth of footage using both and compare the 2.

[auenf]

Originally posted by Jeff D
But, before I get started, auenf stated 16:9 has a virtual horizontal resoultion of 848, is this correct? I learned it as 853, only a difference of 5 but, could make a difference on a 2.35:1 movie. =)

http://www9.brinkster.com/sportschook/?alt.jpg

well, something around there, but 848x480 (and 1024x576) are resolutions that the ati radeon drivers provide for you, and are divisble by 16, where 85x isnt ;)

Enf...

[slk001]

Don't set progressive sequences. For normal television, this is not appropriate (they are interlaced sequences).

[Jeff D]

Thanks guys. I've been learning a lot in the past few days. I had a basic understanding of all this before, I'm now pretty comfortable with most of this.

Originally posted by slk001
Don't set progressive sequences. For normal television, this is not appropriate (they are interlaced sequences).

Scenarist seemed to choke on files with progressive sequences on, I had been leaving them alone. =)

Current to convert 24fps source to playback for dvd run pulldown.exe with the following:
16:9 aspect, 29.97 fps, dropframes TRUE, frames PROGRESSIVE


Dealing with the Convolution3d filter, I was thinking of adding it in before the decimate (this didn't work well). Working with more frames seemed like a good choice, longer, slower, bigger in size, but it really shouldn't make much of a difference, right?

Any suggestions on running convolution pre or post the bicubic resize and adding of the frames?

[Jeff D]

slk001,

I'm not planning on using this on a "normal" tv. It's a progressive HD display, but I don't think this flag is important. I just want to understand. =)


From some ISO MPEG Tech Doc, Looks like I want this flag to be set:

progressive_sequence -- When set to ‘1’ the coded video sequence contains only progressive frame-pictures. When progressive_sequence is set to ‘0’ the coded video sequence may contain both frame-pictures and field-pictures, and frame-picture may be progressive or interlaced frames.

[slk001]

My understanding of the PROGRESSIVE_SEQUENCE setting is that the video (obviously FRAME BASED (FILM) video) is sent to the display as FRAMES (not normal interlaced fields). HDTV has modes that can handle this PROGRESSIVE SEQUENCE - normal TV does not.

[john@sympatico.ca]

I'm reading this thread for interest only. I don't have a LD player + most of it is over my head, so maybe it was explained or is obvious, but I was wondering: How is the LD data copied to the computer? Can you get a direct digital connection, or does it have to go from LD player output (analog), to video capture input on the PC (D to A to D)?

[Jeff D]

John, capture cards via analog video out, svideo or composite, audio is a different beast, you can capture analog or digital.