VHS to DVD help

[diegope]

Hello.

I'm converting my vhs tapes to dvd, I'm happy with the results as far as the bitrate is 6000 VBR or above.
The problem is that now I want to convert my tapes of football games which are about 4 hours each. I would like to fit each game in a double layer dvdr, which means that I should lower the bitrate to 4000 vbr or 5000 vbr, but the conversion is not good. Pixels are quite visible on the edges of the players and the yard lines.
My usual conversion process is:
I have a capture card with a BT878A chip, my pc is a Pentium4 3.2 GHZ with !GB ram and 160 GB hard drive.
The best capture quality I have achieved so far is with Virtual VCR, I capture at 704x576 (I'm in Pal land), Virtual VCR doesnt let capture at 720x576 because lines get across the screen (I'll deal with that some day...).
Then I encode the avi file to an mpeg2 file resized to 720x576 and author the dvd.

I just can´t get an acceptable video quality under 6000 VBR. I've tried 7 or 8 different capture programs testing different resolutions (VCD, SVCD..) and qualities (YUY2, RGB24..).

I've tested different encoders like tmpgenc plus, CCE, but no way, I just cant get rid of those pixels juggling around players and field lines.

I've read the analog capture guide and several threads of this forum so I got plenty of information, I'm not asking for a solution to this problem.

I've read with special attention chapter 4 and 5 (about frame sizes) of the analog capture guide.My question is: ¿is there any relation between frame sizes and my problem?
In other words: ¿Can my problem be solved with correct frame size capture and correct frame resizing?

I've also read carefully section 7 about postprocessing, ¿is there any kind of filter that can minimize pixelation?

Basically all I want to know is if it is worth to test all the frame size and postprocessing of the capture guide to improve the video quality, or if there is really nothing I can do about it due to the capture card limitations.

I know that all I have to do is split the game in two and record to 2 double layer dvdr at a 8000 VBR bitrate, but before I would really like to try to fit them in only one DL dvdr.

Thank you.

[Guest]

Are you encoding in interlaced mode in the MPEG2 encoder? You should be.

[diegope]

Yes, I'm encoding interlaced.

Thank you.

[Guest]

"pixels juggling around" and "pixelization" are imprecise terms. Please post a full-sized screenshot of the effect that bothers you.

[Chainmax]

Yeah, they are quite imprecise. I am betting on dotcrawl and mosquito noise, but it could be other things, so please do post a screenshot or a few frames of the capture.

[diegope]

Sorry for the poor explanation. I live in Spain and right now its 9.30 am and I have to go to work, as soon as I get to my home this night I will post a frame or two. Anyway I will try to explain a bit better.

I think its dot crawling ,squares or blocks appear outlining the players and yard lines, when the camera or players move, these blocks keep changing their color tone and it seems like these blocks crawl.

The lower the bitrate, the more visible it is, it gets even worse at lower resolutions conversions like VCD and SVCD.

Thank you.

[Pyscrow]

4 hours for a football game? cant you cut out the adds and commentator waffle?

[diegope]

There are no ads on football games on spanish tv, and time outs are filled with clips of historic plays and I would like to keep them. If I cant solve this problem I will have to cut the video or split it, but anyway, what I really want to know is which is the nature of my problem so I can concentrate on testing on capture quality or postprocessing the avi file or whatever.
Thanks.

[diegope]

Well, I have attached two zoomed frames of the mpeg2 converted file.
As you can see, players and field lines are outlined by squares or blocks, that change colors and seem to move even in static scenes.

I have have visited Donald Grafts web and I have seen some filters that correct effects similar to this, so I suppose that its noise, or mosquito noise.

I have aldo detected two more artifacts, one is color bleeding, I have found plenty information about how to correct it but anyway it doesnt really bother me.

Please look carefully at mpg3 image, if you look at the left side of the players you will see thar they are outlined by some sort of transparent line, this artifact is in the source tape and in all my videotapes, I suppose that it must be caused by the tv signal. I havent found any information about this artifact and I dont know if it can be corrected.

It doesnt really bother me either, but if I have to filter to correct the avi to eliminate the noise, I would also try to correct the color bleeding and the transparent line.

Thank you.

[diegope]

Please, I would really appreciate if anybody can help me a little bit.
All I want to know is what kind of filtersa can work to try to eliminate the noise.¿should I sharpen or smooth the video?
The captured avi file looks fine, the noise appears in the converted mpeg2 file, ¿is this important?

Thank you very much.

[FlimsyFeet]

I'll try and help, but bear in mind I'm by no means an expert in any of this.

When you said you were from Spain and trying to encode a football game, I expected real football, not... er, anyway, that's another discussion! :D

First thing, you don't need to do any resizing, because 704x576 is valid for putting on a DVD. But I don't think that's the problem.

You say the AVI is OK, I assume you're capturing with huffyuv? It might help to post a frame grab from the capture file before encoding.

I think what you're seeing is MPEG compression artefacting. To reduce them, you need to make the file more compressible. I think therefore you should look at some kind of noise reduction post-processing filter before encoding.

[diegope]

Thank you very much.
After reading so much information about post process filters, I just dont know where to start, I'll test all the noise and smoothing filters and see what happens.
Thanks a lot.

[JeremyIrons]

You might want to try keeping the video in its original capture size to see if this corrects your problem (and....make sure you are getting at least 150-200kbytes per frame of raw captured video before applying any compression). Some programs that resize have crappy stretch or interpolation algorithms, like simply repeating pixels. When MPEG-2 compression takes place using the required 4:2:0 colour space, the interpretation of the colour component when the pixels are sub sampled may be the culprit. The encode process undergoes several steps to yield a compressed video stream, and video that has unnaturally repeated patterns in high contrast or detailed scenes will show up as unnecessary noise.

Mosquito noise for example should only result because of one or several of the following conditions: 1-a poor encode algorithm; --ie: any one or more of its components are badly implemented , 2-too low an encode rate for the video content being compressed, 3-a badly implemented decoder, 4-poor input analog filtering.

Another aspect you may want to check, is that the odd\even fields switch if you have one, be set to odd when capturing (NTSC). You may want to experiment in PAL land which setting has the most effect.

[diegope]

Thanks for the information.
I dont think resizing has something to do with it. Before posting my message I tested for more than 2 months all kinds of capture sizes and resolutions.

Maybe raw capture quality is the problem. At 704x576 with huffyuv in yuy2 mode I get 600 MB per minute (10 MB per second).

Anyway, I've finnaly been able to solve the problem. Just when I was about to start testing noise filters with Virtualdub and Avisynth, I thought about one last attempt with my usual software...and it worked.

When I dont need to convert below 6000 VBR I usually use Ulead Videostudio to encode because it seems that somehow it filters the video noise and enhances the color without loosing detail.

So I thought about converting the raw avi from VirtualVCR to another avi file through videostudio, and it worked.

The rest of the miracle is made with procoder, I encoded the same avi file converted with Videostudio 3 times with procoder at 3 bitrates, 5000, 4500 and 4000. I just couldnt believe it!! I just couldnt see any video quality difference between the 3 clips!!
The noise has been eliminated ( I have to stick my nose to the TV screen to see the noise, true.)
I really cant see any difference between the original source and the converted Mpeg2 file. Amazing.

I hope this information is useful for those sports fans out there.

Thanks.