View previous topic :: View next topic |
Author |
Message |
xo Site Admin
![](images/avatars/11353719634059574e7636f.gif)
Joined: 09 Feb 2002 Posts: 466 Location: Los Angeles [comcast]
|
Posted: Thu Jun 19, 2003 10:40 pm Post subject: SVCD/MPEG2 encoding tips? |
|
|
I'm at my wits end. I'm trying to make an SVCD of some home video footage I shot on DV and I'm having a hard time getting the quality I want. My goal is the same quality that I've seen in other people's DVD-to-SVCD conversions wherein I saw *no* artifacting at all - no banding in flat areas, no blocking in fast motion scenes, no edge artifacts.
Given that my source is at a higher bitrate than DVD and that my footage is quite short (30 minutes total), I figured I should be able to match that quality within the bounds of the SVCD spec.
I've got it pretty good, but there are a couple of scenes where I see mosquitoing (dancing pixels at edges, whatever that's called) and I want to get rid of it.
I've been playing with TMPGEnc and Cinema Craft Encoder, they're pretty damn close in quality, though CCE kicks butt in speed. I've tried assorted options and currently, the best I have is with CCE in CBR at 2520. I figure since my footage is short, CBR at the max bitrate according to spec would yield the best quality.
Still, I've tried up to 9-pass VBR (good, still have mosquitos). I tried setting the quantizer matrix to "low bitrate" (blurry video). Played around with the 4 primary quality controls a little bit (either made it blurry or wasn't any better than the CBR reference encode).
I've spent hours and hours reading and researching to not much avail. Anyone have some "good" settings they use for either TMPGEnc or CCE they can share? This is natural video, not CG/animation, so something tuned for home video type stuff would be great. How are these guys getting such amazing quality in their DVD rips (where it looks superb on a crappy player that does little, if any, smoothing)? Am I chasing a pipe dream?
-xo |
|
Back to top |
|
![](templates/subSilver/images/spacer.gif) |
LeonMcNichol
![](images/avatars/042e73dd3d8117e996f9e.jpg)
Joined: 12 Sep 2002 Posts: 38
|
Posted: Fri Jun 20, 2003 12:37 pm Post subject: |
|
|
It may just be, those scenes need a higher bitrate than the svcd specs. You could always try to higher your bitrate a bit and see if your DVD player will play them. You could also try a CVD or half-dvd, where your video is 352x480 instead of the usual 480x480. That'll give less area to encode and give more room for bitrate. Again, see if your dvd player will play it. Another is, since the SVCD calls for total bitrate not to exceed a certain amount, you could also try lowering 224kbps to like 160 and give a little more room for bitrate. (128 could also work, but that kind of kills some good quality.) And once more, if your DVD player will allow it.
When it comes to encoding, CBR is probably not the way you want for this. VBR is probably what you want to shoot for. You could also try using filters in your avs script.
On a final note, if you are just going to play this back on a DVD player using an analog TV, then I wouldn't bother with this. For the most part, you probably couldn't tell the difference. Analog TVs are very forgiving when it comes to encoding, where digital tvs/monitors are not. Burn it to a CD-RW (if your dvd player plays cd-rws) and see if it looks fine on your analog TV. Even professional DVDs have been known to not be 'perfect'. My steel angel kurumi dvds, I can see artifacting around the edges of lines and such as well as poor quality in some high motion scenes. This of course doesn't show up unless I really really look at my analog tv.
http://www.ravensgarage.com/~leonsite/images/kurumi.jpg
Any jpg artifacting is only slightly minor. What you see, is the DVD. I would have gone png, but that's like 550KB!
Last edited by LeonMcNichol on Sat Jun 21, 2003 5:37 pm; edited 1 time in total |
|
Back to top |
|
![](templates/subSilver/images/spacer.gif) |
Neuralblastoma
![](images/avatars/1843707067413774a0c79f1.jpg)
Joined: 19 Feb 2002 Posts: 109 Location: Ottawa Ontario Canada
|
Posted: Fri Jun 20, 2003 7:38 pm Post subject: |
|
|
VBR is the best way to go, but 4 passes should be your limit. Anything after 4 will not yield any better results. Maximum quality is often achieved by the 3rd pass.
For the mosquito video noise around the edges, use the Undot filter in the AVS script. http://www.avisynth.org/~warpenterprises/
Are you encoding at 29.97fps? If you use IVTC or telecide()/decimate(cycle=5) in the AVS script so you encode at 23.97fps, you will have more space for bit allocation.
An SVCD has to play at 29.97fps though so you need 3:2 pulldown.
If you want to use CCE(which doesn't support 3:2 pulldown) use Pulldown.exe http://home.satx.rr.com/vcdhelp/tools/pulldown.exe after encoding.
TMPGEnc supports 3:2 pulldown. |
|
Back to top |
|
![](templates/subSilver/images/spacer.gif) |
Neuralblastoma
![](images/avatars/1843707067413774a0c79f1.jpg)
Joined: 19 Feb 2002 Posts: 109 Location: Ottawa Ontario Canada
|
|
Back to top |
|
![](templates/subSilver/images/spacer.gif) |
LeonMcNichol
![](images/avatars/042e73dd3d8117e996f9e.jpg)
Joined: 12 Sep 2002 Posts: 38
|
Posted: Fri Jun 20, 2003 11:10 pm Post subject: |
|
|
I agree with neural, converting it to film with pulldown is a good idea. It's less frames for equal amount of bitrate. I also agree about using avs script and using filters through it, though, normally, I don't use any. (Computer is not fast enough. IVTC is slow enough. )
Few other things, you could try bilinear resize, which slightly blurs your video. As I mentioned above, lowering the bitrate, but also trying joint stereo. (Encoded by besweet to get best quality and muxed later with bbmpeg.) I understand when it comes to audio, some people like it to be good, just when you get to high, you overdo it and can't tell the difference. (I hate it when I run into someone who put a 360kbps mp3 in an avi file and the video looks like crap. >< Why must they do that? I mean really, give that little more bitrate to the video.) |
|
Back to top |
|
![](templates/subSilver/images/spacer.gif) |
xo Site Admin
![](images/avatars/11353719634059574e7636f.gif)
Joined: 09 Feb 2002 Posts: 466 Location: Los Angeles [comcast]
|
Posted: Sat Jun 21, 2003 2:05 am Post subject: |
|
|
Thanks for the feedback, guys- I have some new things to try out.
Just to expand on the details of how I've been doing things: I'm capturing DV in Ulead's VideoStudio. For TMPGEnc, I can just use the avis for input since it does the scaling. For CCE, I'm using AviSynth to rescale (BilinearResize(480,480)). I tried bicubic, it worsened the mosquitoing and the video looked harsher. The footage is DV - interlaced 29.97.
All my testing is done on actual TV display (analog) off my standalone player - I don't trust extrapolating results off computer displays, especially since player software have filtering/smoothing algorithms that don't accurately reflect what it'll look like on TV. The artifacts that I described are there in the final TV output and noticeable to me. Of course, I'm looking at it with a very critical eye, but I just can't shake the idea that if others can do DVD-to-SVCD rips that are artifact free, I should be able to do so as well.
I'm still not sure about the CBR/VBR thing. It seems that since my footage is short, I would want to be as close to maximum as possible. Originally, I had it set to 2500 average/2520 max, but the more I thought about it, it seemed that the closer the average was to max, it would simply become CBR anyway. Does a VBR done at, say, 2520 average/2520 max differ from just plain 2520 CBR? I mean, it seems to me that it's not variable anymore...
As for testing out-of-spec methods - unfortunately, this is a video I plan to distribute to friends and family with a variety of different hardware so it'd be difficult for me to test compatiblilty and I want to be sure it's going to be playable (they all play SVCDs), so I want to stay in-spec.
Regarding the IVTC-pulldown suggestion: could you describe the technical theory behind this? I can understand that going from 29.97 to 23.97 gives you more room for bitrate. But then wouldn't going back to 29.97 via 3:2 pulldown take you right back to where you were anyway (so your video is now at a higher bitrate than you want to be)?
Beyond the theory, will this have an effect on the quality? Another issue I've been grappling with is keeping the video sharp and not losing the sharpness in any of the different encodes I've tried.
Thanks again,
-xo |
|
Back to top |
|
![](templates/subSilver/images/spacer.gif) |
LeonMcNichol
![](images/avatars/042e73dd3d8117e996f9e.jpg)
Joined: 12 Sep 2002 Posts: 38
|
Posted: Sat Jun 21, 2003 5:35 pm Post subject: |
|
|
xo wrote: | Thanks for the feedback, guys- I have some new things to try out.
Just to expand on the details of how I've been doing things: I'm capturing DV in Ulead's VideoStudio. For TMPGEnc, I can just use the avis for input since it does the scaling. For CCE, I'm using AviSynth to rescale (BilinearResize(480,480)). I tried bicubic, it worsened the mosquitoing and the video looked harsher. The footage is DV - interlaced 29.97.
All my testing is done on actual TV display (analog) off my standalone player - I don't trust extrapolating results off computer displays, especially since player software have filtering/smoothing algorithms that don't accurately reflect what it'll look like on TV. The artifacts that I described are there in the final TV output and noticeable to me. Of course, I'm looking at it with a very critical eye, but I just can't shake the idea that if others can do DVD-to-SVCD rips that are artifact free, I should be able to do so as well.
I'm still not sure about the CBR/VBR thing. It seems that since my footage is short, I would want to be as close to maximum as possible. Originally, I had it set to 2500 average/2520 max, but the more I thought about it, it seemed that the closer the average was to max, it would simply become CBR anyway. Does a VBR done at, say, 2520 average/2520 max differ from just plain 2520 CBR? I mean, it seems to me that it's not variable anymore...
As for testing out-of-spec methods - unfortunately, this is a video I plan to distribute to friends and family with a variety of different hardware so it'd be difficult for me to test compatiblilty and I want to be sure it's going to be playable (they all play SVCDs), so I want to stay in-spec.
Regarding the IVTC-pulldown suggestion: could you describe the technical theory behind this? I can understand that going from 29.97 to 23.97 gives you more room for bitrate. But then wouldn't going back to 29.97 via 3:2 pulldown take you right back to where you were anyway (so your video is now at a higher bitrate than you want to be)?
Beyond the theory, will this have an effect on the quality? Another issue I've been grappling with is keeping the video sharp and not losing the sharpness in any of the different encodes I've tried.
Thanks again,
-xo |
Capturing... I don't know much about it, but are you resizing from a lower size than 480x480? How about capturing at 480x480? That way, you don't have to resize larger. (Unless you are resizing from 720/640x480, then nevermind.) The reason why bicubic increases the artifacting, is it sharpens. You could try 0.333,0.333 and get a similar effect as bilinear, but still sharpens.
As for VBR, no encoder can keep video within the specs of SVCD compliance and actually has spiked spots. Because DVD players play DVDs which are a lot larger bitrate than an SVCD, you can never really tell. Though some of the older ones do sometimes lag at those peaks. So those peaks sometimes get helped when you do VBR, though it makes it take a lot longer.
Pulldown does not in fact encode again. It doubles certain frames to make it 29.970fps. Sort of adds interlacing to it. It also doesn't change the bitrate. You end up with the same size file you began with. (there abouts. sometimes a few k difference which doesn't bother) |
|
Back to top |
|
![](templates/subSilver/images/spacer.gif) |
Neuralblastoma
![](images/avatars/1843707067413774a0c79f1.jpg)
Joined: 19 Feb 2002 Posts: 109 Location: Ottawa Ontario Canada
|
Posted: Sat Jun 21, 2003 5:41 pm Post subject: |
|
|
xo wrote: | Regarding the IVTC-pulldown suggestion: could you describe the technical theory behind this? I can understand that going from 29.97 to 23.97 gives you more room for bitrate. But then wouldn't going back to 29.97 via 3:2 pulldown take you right back to where you were anyway (so your video is now at a higher bitrate than you want to be)?
Beyond the theory, will this have an effect on the quality?
-xo |
23.97fps is better than 29.97. Film is the best.
For detailed 3:2 pulldown explainations check out these links:
http://www.projectorpeople.com/tutorials/pulldown-3.asp
This is one is super technical and I haven't read all of it:
http://www.dvdfile.com/news/special_report/production_a_z/3_2_pulldown.htm |
|
Back to top |
|
![](templates/subSilver/images/spacer.gif) |
xo Site Admin
![](images/avatars/11353719634059574e7636f.gif)
Joined: 09 Feb 2002 Posts: 466 Location: Los Angeles [comcast]
|
Posted: Sat Jun 21, 2003 11:45 pm Post subject: |
|
|
Thanks again for the help, guys. More stuff to play around with.
Just to answer one question: the footage is DV at 720x480. The AviSynth docs suggested bilinear and I've mostly stuck to that, but tried bicubic in my desperation.
-xo |
|
Back to top |
|
![](templates/subSilver/images/spacer.gif) |
|
|
You can post new topics in this forum You can reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
Powered by phpBB © 2001, 2005 phpBB Group
|