Capturing video, then converting to VCD, why loss quality?

[buzzinhornets]

I captured a test video with VirtualDub and the PicVideo codec thru my ATI Rage FURY Pro 32mb VIVO card and the .avi file looked respectable.... then I frameserved it to TMPGENC mpg encoder using the default ntsc vcd template, the resulting mpeg was crap... super blocky, which is strange cause TMPGENC usually gives me good-looking vcds? What could be the problem here?

Here is pic of my result... not how blocky her arm is...

http://www.thecomputermechanics.com/forums/showthread.php?s=&threadid=23703

[buzzinhornets]

oh yeah, if you have a solution post it on this forum... i wish i could post screenshots here...

[pyrosecht]

what quality of motion search are you using?

[buzzinhornets]

Ummm... what setting in tmpgenc are you talking about... I just use the default NTSC VCD template...?

[pyrosecht]

Go into TMPGenc and load the defatult VCD template. Then click the settings button in the bottom right hand corner. This will pull up a window with several tabs. Click on Video, most of it will be greyed out but at the bottom you will see an option called motion search precision. My guess is you have it set on normal which is what is causing the blocks because they are caused by movement in the video. Change it to at least high. I personally use highest because most of my stuff has a lot of action. It will add some time to your encodes but it will look a lot better when you're done. You might also want to check this out:

http://www.vcdhelp.com/tmpgencexplained.htm

It tells you what everything on the settings menu does. You might also look into learning batch encoding if you don't know how already so that the encode time won't matter much because you can just set it up to do 5 or 6 files when you go to bed and they'll be done by the time you get up. Let me know if you need anymore help.

[Bud]

You might want to try to set some of the VirturaDub Filters, prior to frameserrve to TMPGEns, I use the resize, deinterlace. I Set my Capture Video Format to 352x480 and they resize to 352x240 after the Capture> I get great results for a VCD.

Bud

[volswagn]

When I started doing VCD/SVCD encoding, I used to get blocks even with high quality setting on motion search precision (even with soften noise block enabled). Only way I could get rid of it was to use VERY high quality setting in TMPGEnc. Yeah, it takes WAY longer, but it just looks so much better and very little blocks. Combine that with 2-pass VBR and I think you'll get the best average quality, even if it takes forever to complete.

[Devnull]

You might want to try to set some of the VirturaDub Filters, prior to frameserrve to TMPGEns, I use the resize, deinterlace. I Set my Capture Video Format to 352x480 and they resize to 352x240 after the Capture> I get great results for a VCD.

Bud

Is there really any point to capturing at a higher resolution and then resizing to a lower one?

[vhelp]

Ok, since you pic didn't do you any justice, I saveas to my desktop
and opend it with paint shop and enlarged 1x to make a fair assessment
of your blocks. I didn't really see too much in the pic but then
again, it's not boucing, I mean, there's no movements, so a small
video would have ben better. So, basically, the pic isn't really
doing any justtice for you.

I have Rage Fury APB (not vivo) and gave up on it. I din't like the
wdm drivers for windows 98. so, I'm curently using my ATI-TV Wonder
for my vfw captures via Satalete.

I don't know what you source is, Cable, Satalate, Digital Cable, etc.
So, that may have something to do with blocks too.

* But, you might want to try and cap at 352x480 as others here sug.
* Then, bring into VD and do your filtering. Then, frameserve to
* tmpg. NO NEED TO DEINTERLACE*** Just make sure you have the
right Field Order in tmpg selected. Also, I have my "Video Source
Type" set to "Interlace" under Advanced tab, and under Video tab,
"non-interlace" set - works for me, for VCD's. And give much less
quality loss due to deinterlacing, by the way, "BLEND" would of ben
your choice if you weren't to succeed with my method above.
* you next step is to do a bunch of trial and errors for your final
encodes. Till you find the one that produces the least blocks.
I'm still trying to find the one that works for me.
So far, all most of my sample captures via Satalette still show
some blocks, but then again, I don't get very aggressive with my
settings. I'm starting to realise that you (I) may have to.

-vhelp

*** I'm eating my own words here on this one. As in the past, I've
said you HAD to de-interlace even at 352x240. Hogwash. You don't
have to deinterlace. Only if you're staying with 352x480, then
you have to deinterlace. ...till I figure out how NOT to at this
res.

[dirk67]

Is there really any point to capturing at a higher resolution and then resizing to a lower one?

I always thought it was neccessary to capture at full vertical resolution and then to deinterlace/resize, but now I'm not so sure any more as stated in:
http://www.vcdhelp.com/phpBB/viewtopic.php?t=77394

[Dave B]

For the best quality, always capture in full vertical resolution.

I read the thread which was posted and someone claimed that he thought a deinterlace filter just throws away half the fields, effectively making the action of capturing both fields redundant. Good de-interlace filters merge the information from both fields, giving the mpeg encoder more information and detail to work with when encoding to VCD.

As was mentioned in the post, people need to experiment with these things. Capture a twenty second clip in 352x240 and then the same clip in 352x480. Run both video files through TMPGEnc on the same settings and compare results. This is what I did (in PAL resolutions) and the high res capture output from TMPGEnc was far better.

[dirk67]

@Dave B:

In the thread mentioned i basically compared the following chains:

1) capture 352x576 -> VDub deinterlace -> VDub resize 352x288 -> TMPGEnc 352x288

2) capture 352x576 -> TMPGEnc 352x288

i.e. the same target resolution.