High motion scenes in SVCDs

[arasaki]

I noticed that when I encoded the same clips in VCD and SVCD, the SVCD shows a lot more blockiness during high motion scenes. Is there a way to reduce this during the encoding process?

[stanwebber]

if i knew how you encoded your test samples i might have suggestions to offer. even knowing the bitrate & nothing else would be a wealth of information in comparison

[arasaki]

Ah, of course

I'm capturing to lossless avi with Virtualdub at 352x480. Running minimal filters (again w/ VDub) on it to crop and remove a little noise (plus, for VCD, I do a resize to 352x240). Then encoding it with TMPEGEnc to VCD at 1,150kbit/sec for VCD, High Quality. For SVCD, I encode with at CBR of 2,250 kbit/sec. Low motion scenes look very nice on SVCD, but high motion scenes are very blocky.

[sush]

I use Ulead Video Studio 4.01 to capture via ADS Pyro Firewire at DV resolution avi. Then I encode using TMpegEnc using the standard SVCD template with High Quality for motion. This gives me wonderful results with almost no blockiness. I did this yesterday and was surprised at the quality!

[TMcD]

arasaki,

I'd recommend using a higher bit rate.

I know you might being trying to fit a certain amount of your movie/TV show, etc. onto 74 min or 80, etc. But, it comes down to using more data.

I don't know how bad the blockiness is, and it would be worse if you were using VBR. I am encoding a XSVCD @ 3500 kbps and it looks really good.

As you know, many factors are invloved and it depends on what you want too (such as all on one CDR versus 2, etc.)

I wasted many a CDR before I went and grabbed a CDRW so that I could make test clips (for testing in my APEX 1500.) I recommend that you grab a CDRW and try various clips at different rates.

Troy

P.S. I think VBR is as believable as snipe hunting.

P.P.S. re: sush - If I am thinking correctly, grabbing from DV is like 3.6M/sec. so he'd have lots of data to create a good image.

[TMcD]

After what I said above, vhelp responded to another post I made. You might want to take a look at it.

https://www.videohelp.com/phpBB/viewtopic.php?topic=69620&forum=3

Remember vhelp is the response I am referring to.

Troy

[x2pacalypsehalx]

rofl how you gonna cap under 480 x 480 when thats the svcd size you should expect shit..

[stanwebber]

if you are capturing from vhs, 352x480 is an optimum resolution(broadcast tv & other formats is a different matter). 2250 is a decent bitrate, but there is vast room for improvement. multipass vbr is the first thing you should try to reduce blockiness in high-motion scenes & still control filesize. in any case more bitrate will be of benefit so you might, as someone suggested, ditch the svcd spec altogether

[arasaki]

yea, it looks like I will have to toss out SVCD standard spec if I want this video to look any better. I've tried unsucessfully to do a 2pass VBR because it took so long. A 45 minute video (352x480) had an ETA of over 20 hours? I am thinking of just going with a CBR of 3500kb/s SXVCD. I'll do a test clip and see if my player can handle it first. How long does a VBR encode usually take for you guys?

P.S. I'm using a PIII-700, 512MB Ram WinME system.

[stanwebber]

if encoding time is preventing you from exploiting multiplass vbr i suggest you switch to cinemacraft. for 3pass vbr(minimal filters) i avg .8 realtime * 3 with a p3 800mhz

[kinneera]

Also keep in mind that TMPGEnc offers poor time estimates unless you let it run for awhile.