How to improve SVCD quality ? (from Digital8 source)

[P_Cherry]

I´m converting AVI (captured from Sony Digital8 through Pinnacle Studio 7) into SVCD (SVCD 2520 kps TMPGEnc Template with High quality motion estimate and 10bits precision) but I´m not satisfied with quality of result on my Sony DVD player. The picture is fluiding, restless, noisy (you must take this with ,,reserve´´ because I´m quality freak).

Do you have any experiences, how to improve quality of final result with this setup ?

P.S.: I can´t try CCE because I have Celeron 500 only.

Thanks very much
Peter

[P_Cherry]

I think at this moment, that classic analog VHS has better quality than digital SVCD. Maybe it´s not so sharp, but you don´t have any pixels and fluiding.

[Abbes]

Did you try to use "Soften Block Noise"?
I'm using 20/15 with DVB-captured source. The picture won't be not so sharp anymore, but you'll have less blocknoise!

[DJRumpy]

With CAM sources, you probably do not have enough bitrate to handle all of the shaking motion. Try CVD (352x480). You can still use the SVCD max bitrates, while using less bitrate for the resolution. It may give you that extra umph that you seem to need. I definately would not use 'soften block noise'. It will only result in a blurry movie.

If you have a tripod, use it. Anyone with any cam experience will tell you that one, but I thought I'd throw it in. It makes for a much cleaner and compressable MPEG.

SVCD is superior to analog VHS when done properly, with good source material. It does not do well, with cam sources with a lot of motion or jitter. It swallows all available bitrate just to compress the scene changes from frame to frame, leaving the image quality suffering.

You can try to help maximize your motion detection, using multipass VBR, with min, max, and avg set to your target bitrate. The additional passes will help the encoder allocate bitrate better than a single pass using CBR.

Last but not least, lower the audio compression, and give the extra bits to the video. Chances are, you cam audio is probably not the best. Don't waste 224kbs on it. Use something lower. Try a few comparisons between the original source, and a lower bitrate (say 96kbs). See where you find your happy spot. Put the extra bits onto the max kbs for the video (every bit counts....)

[P_Cherry]

Thank you for your replays, I have some commentary and question to it:

1) I use maximum CBR 2520 kbps and I dont´t think, that when you use multipass VBR, you will receive better results, because VBR only changes bitrate between minimum (eg. 1000) and maximum (2520) alongside the picture. And how you could receive better result with same bitrate (picture information) ?

2) I also tried CVD (352x576), but it seemed to me maybe little bit worse, than SVCD. The picture remained fluiding and pixelating also. And I think was less sharp (it´s caused by decreasing of resolution, of course).

Peter

[WarpEnterprises]

1) Does your DVD only do up to 2500?

2) Besides using other progs (ProCoder gives excellent results without needs to tweak but is really too slow for your machine) you could try to make a "widescreen" video by cutting off some lines from top and bottom and leaving 16 pixel left and right (overscan) - so you get a smaller motion area with more bits per pixel.

[P_Cherry]

Yaeah, my SONY DVD player has maximum bitrate for SVCD about 2500 kbps, higher bitrates causes ,,freezing´´.

What do you mean with widescreen, I don´t understand to it exactly ?

Peter

[banjazzer]

To get best quality you need to capture through Pinnacle uncompressed through a firewire card. Even then, in my opinion Studio is nowhere near as good as TMPg for encoding to SVCD. What I would do is edit in Studio then save back to your camcorder (if you have DV-in). Recapture with Studio and encode the DV with TMPG.

Although this sounds a bit crazy, my experience with Studio is that after editing there is no option to save to uncompressed format (except back to the camcorder). This may have changed with Studio 8.

[WarpEnterprises]

"widescreen":...
you know: the more black bars visible the better the quality with same bitrate.
You can use this for letterboxing the overscan area (the parts never visible on TV) or going even farther to crop away top and bottom of your video (of course this makes only sense if there is nothing interesting there which often is so).
But it won't get you much advantage.

[P_Cherry]

TO BANJAZZER: Thanks for your time, but you must read, what I wrote. You adviced me the same thing, that I had been doing so far.

TO WARPENTERPRISES: My movies are in anamorphic 16:9 format and that´s why I don´t want to add black frames. And as you said, the advantage will not be so clear from it.

[DJRumpy]

P_Cherry, he is suggesting letterboxing to basically block out the edges of your video. You would lose some information on the top and bottom, but you could then apply that bitrate to the high motion areas of your video. Essentially making a 2.35:1 ratio movie, which has letterboxing, even on widescreen. The benefit, is that the black areas do not require bits, except on a keyframe, when encoding in MPEG. Those spare bits can make a big difference, since the letterboxing can use 10%, 20%, or more of your movie.

To answer your other question regarding multi-pass VBR, you gain benefit from using it, simply because the encoder can better allocate bits from frame to frame. The encoder 'learns' something from each pass. It can then take bits from area of your video that don't need the full 2500 bits, and allocate them to the area's that do, minimizing your bitnoise. I would suggest you try 2-Pass VBR if your using TMPGenc, and3 pass if your using CCE:

Min: 0
Avg:2400
Max:2524

If your player has a problem with the MIN, try 300.

If you don't want anything less than 2524 bits, and you still want the benefits of multipass, then set your min, avg, and max to 2524. You should be aware that not all frames in your video will require that many bits. If you set all of yours to the max, then your simply wasting bits that could improve your video quality in more difficult scenes.

If your CVD looks worse, then the resize your doing is probably the culprit. CVD requires less bitrate than SVCD, and the difference in quality due to the change in resolution should be negligable.

[banjazzer]

TO BANJAZZER: Thanks for your time, but you must read, what I wrote.

My apologies. I did read what you wrote, but I presumed because of the lack of quality you were not capturing as uncompressed DV, but as mpeg, since you did not specify the capture format.

[P_Cherry]

TO BANJAZZER: Ok. Nevermind. I have Studio 7 only, capturing directly to mpeg is possible in studio 8 and higher only.

TO DJRumpy: If I understood good, when encoder goes through the movie on the second time, it knows for example that in the middle of the picture needs more bits than on the top and corrects it. When I say this very simply.

[DJRumpy]

P_Cherry: Yes, you understood correctly.