Telecined tapes - unsatisfactory results

[wylenda]

I've read all the faqs, forum threads and experimented alot but I'm still not satisfied with the DVD's I've created; here's my story.

I'm capping with a JVC S-VHS to a WinTV PVR150 and AC'97 Line In. The episodes were recorded from DVB satellite using S-Video output onto premium S-VHS tapes - result is excellent. I'm using Virtual VCR and HuffYUV codec and I get no dropped frames as long as I'm not doing anything else on the pc while it's recording - or else I get lip sync problems.

Now the problem I'm having is what I could best describe as "motion blur" when someone walks across a room for example and the camera has to pan to follow the character.
I'm using VirtualDubMod with AVI script to crop the bottom 7 lines and then do a Resize (Lanczos). The only other post processing I use is to use Inverse Telecine option in Vdub as per the Capture FAQ. I've used TMPGEnc, CCE and Procoder2 encoders and prefer the results of the former and am very disappointed with the Canopus encoder (halos around edges ,as if something cranked up the sharpness too high !)

I paid for and used the Picvideo Mjpeg codec with settings from 16-19 and I must admit I don't like the results - is this a vfw codec ? I thought it was WDM when I bought the licence.

My last question is to the avs script and Vdub filters - I want to try some avs IVTC filters but I'm not sure where in the script it needs to be (see my script below). Do I even need to IVTC if I'm only watching on DVD ? When I frameserve the clip when does it IVTC ?

Lots of questions I know but I've been working on this issue for quite some time now and finally feel defeated and thought I'd ask for help.

One last thing; on some episodes I edited the commercials out while watching and this left a nasty "rainbow" bar when I un-paused record again and would like to know the best way to get rid of this.(btw is this because this model doesn't have flying erase heads ?) oh before I forget I use 6000 kbps 2 pass VBR to encode the 400 line S-VHS stuff , is this overkill and what should my mininmum and max be in CCE and Canopus ?

thanks and I hope I've explained myself well.

BTW I've confirmed these are telecined broadcasts (3 progressive and 2 interlaced).

AVISource("E:\VIDEO_OUTPUT\capture.avi")
Trim(0,14300) ++ Trim(16572,29368) ++ Trim(35111,47520)
CropBottom(7)
Lanczos4Resize(720,480)
# where do I put IVTC filter ?

[FlimsyFeet]

If you're making a DVD then IVTC is not essential, you can encode as interlaced. But I think it is preferable to IVTC and encode as 23.976 progressive. It's probably more efficient.

The IVTC operation should normally be carried out before any other processing. If you resize before IVTCing, that will cause blurring.

Also, one minor thing, I would suggest AddBorders(0, 0, 0, 7) instaed of Lanczos4Resize(720,480).

[Boulder]

If you're making a DVD then IVTC is not essential, you can encode as interlaced. But I think it is preferable to IVTC and encode as 23.976 progressive. It's probably more efficient.
It's a helluva lot more efficient so it's essential that you IVTC. I would place the IVTC process right after AVISource. Download the Decomb package by Donald Graft, it includes a good tutorial on how to IVTC.

Also, one minor thing, I would suggest AddBorders(0, 0, 0, 7) instaed of Lanczos4Resize(720,480).
Or better yet : use LetterBox instead of CropBottom+LanczosResize.

What do you mean when you say that you don't like PicVideo MJPEG? At Q19 it should look almost the same as a lossless capture. There's also a very good (and free) MJPEG codec included in ffdshow.

[wylenda]

What do you mean when you say that you don't like PicVideo MJPEG? ,<snip>.

I take that back, I should have said Vdub pukes with my avs scripts when I cap with it ... so I stick with Huffyuv for now - strange I know.

I took your advice and here's my new script,

LoadPlugin("C:\Program Files\AviSynth 2.5\plugins\decomb.dll")
AVISource("E:\VIDEO_OUTPUT\capture.avi")
Trim(0,14348) ++ Trim(16623,29417) ++ Trim(35160,47569)
AssumeTFF()
Telecide()
Decimate(cycle=5)
Letterbox(0,7)

I did read Donald Graft's de-interlace tutorial and followed it to a tee, thanks. Now I get a strange stuttering effect in movement. Can this be cause my pc is too slow ? I'm converting from RGB to YUY2 in Huffyuv when I cap in VVCR ask suggested by Mr. Graft.(btw any good tuts on colorspace ?)

I'm not sure about the order in my scripts - is it ok ?
What if I add Tweak() ,this goes at the end ,right ?

Sorry about all the noob 'ish q's.

Thanks for all the help, appreciate it.

Thanks for all your help, having fun experimenting !

[Boulder]

I took your advice and here's my new script,

LoadPlugin("C:\Program Files\AviSynth 2.5\plugins\decomb.dll")
AVISource("E:\VIDEO_OUTPUT\capture.avi")
Trim(0,14348) ++ Trim(16623,29417) ++ Trim(35160,47569)
AssumeTFF()
Telecide()
Decimate(cycle=5)
Letterbox(0,7)

I did read Donald Graft's de-interlace tutorial and followed it to a tee, thanks. Now I get a strange stuttering effect in movement. Can this be cause my pc is too slow ? I'm converting from RGB to YUY2 in Huffyuv when I cap in VVCR ask suggested by Mr. Graft.(btw any good tuts on colorspace ?)

Capture in YUY2, that's probably the native colorspace of your card.

Do you have the stuttering in the encoded file? If so, you either have an incorrect field order or the material cannot be IVTC'd. If you are unsure, you can upload a small, unprocessed sample somewhere and I (or someone else) will have a look. Ten frames or so should be enough.

I'm not sure about the order in my scripts - is it ok ?
What if I add Tweak() ,this goes at the end ,right ?

The order looks ok, and Tweak should be after the IVTC phase, yes.

[wylenda]

Capture in YUY2, that's probably the native colorspace of your card.

I'm not sure how to select colour space in VVCR - the only option I have is RGB24 in the Video tab ->colour format.
I'm a little confused about all this , I apologize. In the Video pin setting I have "HCW2" for color space/compression as the only option , I don't even know what that is , Hauppauge ?
For example if I capture with no compression and then try to open the video in Vdub I get a warning that says "Couldn't locate decompressor for format HCW2(unkown)".

Do you have the stuttering in the encoded file? If so, you either have an incorrect field order or the material cannot be IVTC'd. If you are unsure, you can upload a small, unprocessed sample somewhere and I (or someone else) will have a look. Ten frames or so should be enough.

The stuttering appears in the encoded video file as when I IVTC and blurred when I don't IVTC. In the encoder I have the option to select interlaced and which field order but I'm never sure what to do here so I select interlaced top field first -how do I determine the field order the card is capturing in ?

I've uploaded a 20 frame sample clip as you requested - please have a look and let me know what you think.

pj-sample.avi - 19.91MB (http://www.zshare.net/video/pj-sample-avi.html)

[Guest]

Don't do your trims before IVTC because cutting 3:2 pulled down material can change the field order! That may give you sections where the field order is wrong.

IVTC the whole thing and then do the trim editing.

[Boulder]

Yes, I totally forgot that IVTC should be done first, even before trimming.

However, the sample clip looks like it doesn't have a steady pattern. I got NNNCCNCNNNCCNNNNNCCN where N is a non-combed and C is a combed frame.

What comes to the colorspace, it could be the card that forces you to use RGB24 for some odd reason. Have you tried capturing in MPEG2 (doesn't the PVR-150 have a hardware MPEG2 encoder chip?) or using VirtualDub 1.6.x?

[wylenda]

Well I now see I've posted that sample clip and indeed it is an odd pattern - I've played around with the settings in the decomb plugin trying to teach it how to determine progressive from interlaced frames and it can't lock onto the pattern, why is this ? (I can see the combing when I manually advance in zoom mode) I have proper 3:2 captures that work great with the plugin - thanks Donald Graft.

I'm disappointed that I can't get smooth motion from the irregular clips with or without IVTC.

Couple of problems: With the Picvideo MJPEG captures Vdub reports the wrong fourcc code for this codec (thinks it's XviD) and as a consequence shuts down when I try to do things like Letterbox so I'v been forced to use crop/block in TMPGEnc to accomplish this.

Have you tried capturing in MPEG2 (doesn't the PVR-150 have a hardware MPEG2 encoder chip?) or using VirtualDub 1.6.x?

Yes it does have a h/w mpeg2 encoder but how do I edit the video head switching noise and other things I can do to the avi ? Also the card consistently has audio distortion and lip sync issues. I really like the "hard" way and am very pleased with the results I'm getting.

I'm not sure after I Decimate() to get 23.976 fps , when I encode the video it goes bac kto 29.97 interlaced , doesn't that defeat the whole point of IVTC'ing ? Can my DVD player play the 23.976 progressive clip directly ? If so how can I encode a DVD at this frame rate in TMPGEnc for example ?
Should I leave off the Decimate() bit in my script then ?
Won't this cause a jerkier motion ?

Sorry about all the questions - learning as I go along.
Thanks.

[Boulder]

You can process the MPEG2 captures just like you would process a regular DVD. Open the mpeg file in DGIndex, create a d2v project file and load it in your script with MPEG2Source. Then proceed as you normally would.

From what I can remember, lots of issues have been solved with new driver versions. You could also check www.shspvr.com , there's a good forum there that is concentrated on the Hauppauge PVR cards.

[wylenda]

You can process the MPEG2 captures just like you would process a regular DVD.

I was wondering about decoding and processing mpeg2 captures, thanks for pointing that out ;)

Turns out I already had the DG package when I installed AutoGK for I just copied the dgdecode.dll into the avisynth plugin folder.

I know about that hauppauge board thanks anyway.

I still have to re-encode these back when I'm done processing right ? Something tells me I don't have to ...

Q: I'm still having that annnoying bug in VdubMod reporting the incorrect codec when I use the Picvideo MJPEG codec (not so for Huffyuv), and I can't use things like Letterbox() - strange cause it reports the correct fourcc code when I open the avi directly in VdubMod - the problem only occurs when I open it with the avs script ... has there been a fix for this and anything I could do to remedy ? I want to use VdubMod to fileserve sometimes cause I'd like to use color mill to "enhance" the captures.(I'm saving that for another thread)

Thanks for all your help

[Boulder]

When you load an avs script in VDub, it doesn't know what codec has been used to access the actual video file. It always reports the codec that is used to decode the raw stream (be it YV12, YUY2 etc.) Avisynth provides.

You can use ColorMill in your script, Avisynth supports loading VDub plugins. See the Avisynth documentation for more details. If you are wondering what the settings for ColorMill should be in the script, you can open the video file in VDub, add ColorMill, adjust the settings and save as AVI. Tick the box that says you'll run the job later and proceed. Then you can open the file called virtualdub.jobs found in the VDub folder and get the parameters for ColorMill there. Use those in your script.