CVD aliasing : how to get rid of it ?

[waldok]

Hi,

I really want to move from SVCD to CVD with DVD2SVCD, but it seems I can't get rid of a very visible aliasing (stait effect) on near-vertical diagonals.
I encoded with CCE at CBR 2400 and used bicubic resize (0; 0.6)and Temporal smoother(2,1). I tried both NTSC (352x480) and PAL (352x576) resolutions without better results. I can clearly see a medium-to-strong aliasing on diagonals.

Is this due to 352x576 to 4:3 interpolation ? Is there a way to correct this and get a good picture ? (Any use of sharpen here ?)

Thansk for helping cause I can't uderstand how so many people find CVD great when for me SVCD stills remains far better. (DOn't tell me to stick with svcd, I'm doing this for : a. more bits per pixel at the same bitrate and b:Future DVD-R compatibility.

THanks a lot for your help

PS : Adam, are you the same adam who answered one of my posts in Doom9 forum lately ?

[Guyver]

I've noticed the effect too, but it was only visible on the pc. When i watched it on the tv there was no effect. So maybe it just happens on the pc.

[waldok]

THanks for answering, but I can tell you that this damn aliasing is quite visible on my TV (a 109cm rear projector), even from 4m far away !

Waldok

[SatStorm]

I can only say that when I use Huffyuv to grabb with virtualdub and then I encode with tmpgenc plus 2.56, I don't have this effect.
With Mjpeg PicVideo, I have this aliasing (stait effect) you describe.

I never have this effect with DVD and DVB source

[adam]

Yeah waldok thats me. Ive now read the couple of threads you started on doom9 and this one here but I don't have any kind of answer for you.

Like others I just havent noticed this aliasing effect at all. I have a rear projection tv at my mom's house so next time I'm there I'll try to throw in one of my CVD's and see if I notice it but I don't remember ever seeing it in the past.

I know that as far as this aliasing is concerned bicubic resize would probably be better, but neverthe less I usually use billinear when resizing in only one direction ex: 720x480->352x480. If you havent already try using another resizing method.

Hopefully this is a fixable problem and not just a limitation of your dvd player or tv. How do vcds look? If this were a hardware problem than I would think that it would be just as apparant on vcds.

Oh by the way, my sources are typically ntsc dvds.

[waldok]

Hi adam,

You are pointing out a very interesting thing I didn't think of. I will give VCD a try to check if the aliasing problem is due to a poor interpolation from 352 to full width either by my DVD player or by my TV Set (I could suspect DVD player here).

INteresting question by the way, where does this "picture enlargment" takes place? Is it done at decoding time by the MPEG decoder ? Is it done afterwards, by a specific chip in the DVD player ? Or is it sent as 352 to the tv set (i don't think so).

Who knows about this ?

Waldok

[bbb]

In all the VCD and CVD clips that I've downloaded and encoded, they do show more jaggies than SVCD clips using WinDVD on a computer monitor.

On a standalone dvdplayer using a TV, the jaggies pretty much disappear.

I always assumed that the jaggies are due to the limitations of the horizontal resolution for VCDs and CVDs when enlarged to fit a 4:3 window or to full screen.

The jaggies for CVDs are only noticeable where there are large black outlined diagonal lines or arcs and watched on a computer screen--especially at full screen.

If jaggies on a computer screen can be removed by using different drivers or codecs, I would be quite pleased. However, I doubt that the solution is that simple.

Interesting note, if you capture at 352x480 but encode to 480x480, the jaggies are gone. This is further evidence that CVD's lower horizontal resolution is the cause of the jaggies when the CVD video clip is enlarged. This is one reason to stick with SVCD.

[energy80s]

Interesting note, if you capture at 352x480 but encode to 480x480, the jaggies are gone. This is further evidence that CVD's lower horizontal resolution is the cause of the jaggies when the CVD video clip is enlarged. This is one reason to stick with SVCD.

Wrong. If that were true then your "CVD" capture at 352x480 would exhibit the jagged edges AT ANY SCREEN RESOLUTION!! And would look worse at 480x480. The fact that you are getting rid of the effect by upping the re-encoding resolution means that CVD isn't your problem at all.

[bbb]

My oh my. What a touchy CVD cultist.

I downloaded a SVCD TV episode. According to the person who encoded it, he captured at 352x480 and then encoded the movie at 480x480 resolution. He applied many filters and color correction and encoded the AVI with CCE. I don't have the details as to his resizing process but, his end result at 480x480 did not show any jaggies unlike CVDs. I was quite surprised.

Perhaps a high quality bicubic resizing from 352x480 to 480x480 may eliminate the appearance of jaggies. Someone should try this out with a little clip. My encoding computer blew up so it'll be 1 week before I can test this out.

Meanwhile, I'll try to confirm from the author of the SVCD clip whether he really used 352x480 for capture.

[energy80s]

My oh my. What a touchy CVD cultist.

Damn Right!!

[bbb]

Waldok,

I think you are correct in your initial assessment that CVD's low horiztonal resolution is the main cause of the appearance of poor antialiasing on diagonals that show up prominantly on computer dvd players. The poor antialiasing is even worse with VCDs. They pretty much disappear with SVCDs.

It's too bad CVD doods, pretend that their are no flaws/limitations with CVDs. Yes, I encode to CVDs but I realize that it's a compromise for fewer blocks for less clarity.

I suppose it's only natural for people to be greedy for attention. But, encoding 352x480 has been discussed for at least 1 year. Before, I've avoided it because it looked blurrier than SVCDs. Now, I use it because of the supposed compatibility with DVD recordable discs. I do try to address the blurriness by using sharpening filters but that also increases noise and makes the diagonal edges look jagged.

If any CVD dood tries to tell you that they can encode video material with diagonal edges without showing jaggies (e.g., buildings, car window edges), ask for proof. I doubt they can provide any credible evidence in terms of video clips. If they tell you to use TV not computer monitors, then they have failed to prove their point.

To sum it up: CVD is a good TV format (far better than VCD). However, it's limitations will show up on a computer screen. While CVDs do have DVD compatibility, when HDTVs come standard, CVD resolution will begin to show their flaws on HDTVs just like on a computer monitor.

Whatever happened to the header trick that lets you make SVCDs clips compatible with DVDs? Oh, here it is:
http://www.vcdhelp.com/svcddvdr.htm

[waldok]

Well, I don't have to wait for HDTV to see the jagged effect since I can very clearly see it on my 43" (109cm) Toshiba rear-projector. I never use my computer to watch Svcds or CVDs. At this level of "aliasing", I consider the movie to be nearly unwatchable. For example, I tried CVD max cbr bitrate for a chapter from the movie SPy Kids (PAL region 2). In this chapter, you can see two kids in a store. believe me or not, the stair effect on the kids face edges are really visible from a 4 meter viewing distance...creating a feeling of a very confusepicture cause faces edgae sseem to melt with background pixels in the picture (No I don't do drugs...)

Maybe CVD wouldbe great (considering the "more bits to allocate to fewer pixels" thing) if one could add a specific horizontal antialiasing filter at decoding time, which would probably have to be hardware located in the DVD player itself or in the TV set.

So I think I'll be on to some more experimentations here with CVD (probably for Svideo family videos (resolution should be OK)), but a far as DVD rip is concerned, I'll stick to SVCD for the moment.

By the way, bbb, thanks for pointing out the interesting DVD SVCD header trick.


Waldok

[SatStorm]

It is not the resolution, is the wrong Gop structure. Set PAL gop structure to 15 and don't set "closed Gop" any more.
That way the picture is better and acceptable from DVD authoring progs.

[bbb]

Waldok,

I just realized something (just got my pc running smoothly). Do you encode your clips with interlaced mode?

Some movies are heavily and inconsistently interlaced. Even with a proper inverse telecine (IVTC) using a low threshold for horizontal stripes to remove interlaced lines, some interlacing horizontal lines still remain in the video. I have noticed that Interlacing lines/artifacts makes jaggies on curves and diagonals appear far more visible.

As such, as a default rule, even after IVTC, I deinterlace with Double (aka Blend) in Tmpgenc. I also slightly Sharpen Edges to counteract the slight blurriness resulting from Double (Blend) deinterlacing. Then I encode ALWAYS in progressive mode. This gets rid of ALL interlaced lines. Virtual Dub also has a blend deinterlace filter so you can try this step with CCE and encode in progressive mode.

Diagonals and curves will still show jaggies on a computer screen but they won't be extremely pronounced. They will almost be nonexistant on a TV. And I'm ALMOST certain that the Spy Kids druggy blurriness on the kids faces will disappear.


SatStorm,

GOP SMOP SPAM WAM. Sorry, I don't buy into Kwag's hype regarding excessive long GOP tweaking and CQ speed+quality reliance. I downloaded his latest CVD clip (oh excuse me, I mean his SKVCD clip) and it looks very nice and I can still spot CVD jaggies.

Then I encode my own CVD clip using standard SVCD template but changing the resolution to CVD and using ITVC, 2-pass VBR (900 minimum, 1800 average, 2450 max), Max Frames per GOP: 16 (for DVD compatibility), and progressive encoding.

My result: just as nice and I can still spot CVD jaggies. Conclusion: Kwag has wasted too much time with GOP tweaking. By using a smaller DVD compatible GOP structure, I was able to achieve the same nice quality and small (and also predictable) file size from a DVD movie. Also, by using a smaller DVD compatible GOP structure, you are far more likely to obtain DVD compatibility for the future.

Ah yes, the future ... when a dominant DVD standard arises from the dust ... where we can transfer multiple 2 (or more) CVD disc movies onto 1 DVD.

[kwag]

Conclusion: Kwag has wasted too much time with GOP tweaking. By using a smaller DVD compatible GOP structure, I was able to achieve the same nice quality and small (and also predictable) file size from a DVD movie. Also, by using a smaller DVD compatible GOP structure, you are far more likely to obtain DVD compatibility for the future.

Ah yes, the future ... when a dominant DVD standard arises from the dust ... where we can transfer multiple 2 (or more) CVD disc movies onto 1 DVD.

You did indeed achieve the same quality, and I did mention this on other posts. What you will never achieve with that quality is 60+ minutes on a single 80 minute CD-R, because or our own Q.Matrix. It's not just the GOP. No matter what encoder you use, or what encoding method you try.
As far as DVD compatibility, I really don't care. If I want DVD compatibility, I'll encode with KDVD templates, and burn with any DVD authoring program to DVD (+-)R and put either 4+ hours or 7+ hours with the "Half D1" template, and play in every DVD in the planet.
But if you insist on puting SKVCD on a DVD, just change the audio to 48khz and the MAX frames per GOP to 15 ( or 18 for NTSC ) and there you have it. Full DVD compliant MPEG, still compressing better than any other encoder, as long as you use our Q. Matrix. You'll still be able to fit around 50+ minutes on a CD. It's that simple, so I don't consider it a waste of time


-kwag

[bbb]

I CAN fit 50+ (around 55) minutes on MY DVD compatible CVD template too using a 1800 average bitrate for 2-pass VBR. These were the settings use to compare my sample to yours (Kate and Leopold). Both mine and yours looked equally good.

Why bother messing with Q or GOPs if it makes no difference, other than breaking DVD compatibility?

As for putting a FULL movie on a cdr, I have never seen a sample (including your Matrix samples) that looks as good as a 50-55 minute CVD. Once you drop under 1600 bitrate for CVD resolution, blocks will show up unless it's a very slow moving movie or it will look great only for slow motion/transition scenes. Overall, <1600 bitrate clip may still look decent (better than VCD) but does not arise to the level of a great looking CVD. This has been the case for every clip that I create or have downloaded, regardless of the hype.

Video quality on a CVD is basically determined by using a clean, sharp, high resolution source and adequate bitrate, assuming that you use Tmpgenc or CCE to encode. CVDs/SVCDs will not look their best once you pass the 60 minute mark per cdr. No amount of tweaking can fit more than 60 minutes of CVD video per cdr without yielding to more blocks.

Perhaps in the future a revolutionary encoder will be developed that can utilize VCD bitrates to produce block free, crystal clear video on CVDs. As for now, it's wishful thinking.

[kwag]

@bbb,
About 99% of people doing SVCD don't care about DVD compatibility. They care about SVCD compatibility. Why?, because we'll be re-encoding our movies in the future again, when HD-DVD's are standard, and DVD's are obsolete
So if someone want's the best quality/space, while still maintaining SVCD quality, as you have already agreed!, then stick to SKVCD, as everyone will be able to stick a full movie in two CD's. Not three, as with a regular SVCD. That "Just one more" CD-R is just "One too many" for me ( and many people! )

-kwag

[bbb]

I realize that 99% no make that 100% of your followers (a.k.a. vcdhelp newcomers) will listen to your every decree. It must be fun being a living diety. However, I'm not looking for that kind of lifestyle.

I just want people to know the consequences of their current encoding methods.

People may not think so but they should prepare for the future, when DVD recordables will be standardized, multispeed DVD writers will drop under $100, DVD blanks will cost under $1. People should look to compatible standards so they don't have to spend thousands of hours reencoding and degrading the video quality of their video (which could include treasured moments) into a dvd format. This future may be 2-3 years off but it will be here before we know it.

If you deinterlace and encode your video into progressive DVD compatible format, you will not need to reencode (and lose quality) in the future for DVD and HDDVD compatibility (HDDVD players will play DVDs).

This is why you should stick to future compatible standards. I'm sure some of your followers will agree if you let them have free thought without fear of banishment.

[kwag]

The best thing in all this stuff we're doing is called "Choices" 8)
So if someone wants to stick to DVD compliance, they can. If they don't care, and want absolute best quality in another format, they can too!.
So it's the user's choice which way to go. I personally will be very happy to re-encode new HD-DVD releases to whatever format is current in the future. But hopefully, my current encodes will still play, as most DVD players are backwards compatible.
Heck, who knows, at the speed of technological developments who knows what we'll be playing in 5 years

Just like the Vacuum Tube. You think it's obsolete today? Ha Ha take a look again at this new motherboard. This is wild http://club.aopen.com.tw/activity/tube/en/default.htm I want one

-kwag

[adam]

stick to SKVCD, as everyone will be able to stick a full movie in two CD's. Not three, as with a regular SVCD.

Kwag SVCD can be encoded at as low a bitrate as you want. Obviously you know this but I do think that it is a common misconseption that SVCD necessarily needs to be 3 cdrs to look good. You can fit up to 60 mins per disk in SVCD format in very good quality. Most movies are not longer than 2 hrs so most movies will still fit on 2 disks in SVCD format. Even for longer movies its not unreasonable to still use 2 disks, not any more unreasonable than if you used your SKVCD template.

There are other Q Matrices available on the internet, many of which are actually used by the major production studio's. I've experiemented with all of them, including Kwag's, and personally I see very little difference in any of them as far as the quality of the output. The same goes for GOP modifying. It can help that's for sure but, assuming everything else in your process is sound, it still all comes down to your bits per pixel and the distribution of those bits. Kwag's template has some interesting modifications which can take a little bit of the edge off of low bitrate encodes, but I don't see any real quality improvement. If you want to fit 60+ mins onto the disk then simply lower the bitrate, simple as that.

And just for the record I still prefer the quality of my 2 cd SVCD's to a 2 cd SKVCD (or whatever you wanna call it.)

[Martyn1980]

Im among the people who can see the aliasing on CVD on a TV and we're talking a 21" TV, and i stopped using the format because of this and the fact the colour looks wrong. VCD and DVD are proper TV resolutions SVCD and more so CVD need to be stretched to fit a tv properly and i dont know how these people with massive TV's dont notice it.

As for getting a film on 2 CDs, Kwag it sounds like you think everyone not using your templates needs 3 CD's, which isn't true, im actually messing with encoding using some of your new templates and old ones, going to send them back to my brother who has my Toshiba SD210e and see if they work as your site says and hopefully they will look good.

How did this post turn into another encoding argument, when it was a anti-aliasing question.

[kwag]

How did this post turn into another encoding argument, when it was a anti-aliasing question.

I don't know. But I guess it has to do with negative karma on some people
Specially when you find comments on other sites, like the third post on the following link: http://www.isonews.com/forums/showthread.php?s=2d21e0799901023dcd9cba819db86769&threadid=88326

Here's a quote of the message:
"Its not a new format its just a non-standard SVCD.
All it is, is a template for TMPGenc that has longer GOP's (common encoding technique) and a modified Q Matrix which arguably makes very little difference. If you want a really good Q Matrix get the Andres one instead. There have been similar templates available on the internet for years.

If you want to test this template I'd do an encode myself and not trust those samples. They use alot of noise filtering to offset the low bitrate and the image is too soft for my taste. Play it back next to a standard SVCD and you'll see what I mean, and see what you're missing."

So you see, this person is not only misleading others, he's also a liar. Because all the KVCD plus and SKVCD samples were encoded from VOB's WITHOUT any kind of filtering. But personally, I don't care, because I've only had 2 or 3 negative comments on SKVCD, and over a thousand kudos. So those kind of comments are just, well, you guess


-kwag

[bbb]

How did this post turn into another encoding argument, when it was a anti-aliasing question.

It's hard to keep new age religious fanatics out.

So, getting back to topic, I have confirmed that CVD's resolution is the source of the problem.

Using a VOB (dvd resolution) and an AVI (640x480), I encoded it into CVD and SVCD resolution. The SVCD clips had no (or very imperceptable) jaggies when played on WinDVD. The CVD clips showed jaggies on diagonals and arcs.

Then I used Virtual Dub to resize my AVI to 352x480. I frameserved and encoded it with Tmpgenc using CVD (352x480) resolution. Result: Jaggies appeared in diagonals and arcs.

Again, I used Virtual Dub to resize my AVI to 352x480. I frameserved and encoded it with Tmpgenc using SVCD (480x480) resolution. Result: Jaggies disappeared.

Conclusion: CVD resolution is too low and will show jaggies in diagonals.

But for practical purposes, CVD is decent for VHS captures to be played on standalone dvdplayers.

[adam]

Kwag I have never seen a single test or sample that you have made that you didn't claim to use noise filtering on. If your most recent samples do not use noise filtering than I appologize, but even still the image is way too soft for my taste. Why did you all of the sudden decide not to use filtering anymore? In any case, I have yet to see a sample of yours that was actually representative of your template. I don't see anything wrong with telling people to try the template themself rather than trusting your samples. Also I don't see how telling people to think for themselves can be considered misleading them.

I don't understand how my opinion can be considered a lie either. I'm not impressed by your templates or your samples and I have no problem stating it, on any forum. I will continue to use what settings work for me and you will continue to peddle your settings on every forum available.

[kwag]

Conclusion: CVD resolution is too low and will show jaggies in diagonals.

But for practical purposes, CVD is decent for VHS captures to be played on standalone dvdplayers.

Not to mention that we'll have to live with 352x480 resolution, as it is a standard DVD resolution ( Half D1 ), and every DVD player supports it.
DVD Recorders like the DMR-E20 use this resolutions for one of it's recording modes. And it plays perfectly on every DVD player.
Not so for 480x480

-kwag

[Martyn1980]

I dont really care about if its a valid DVD resolution as i would be ashamed to stick it on a DVD, people dont really expect much of VCD's SVCD's and i like surprising them at the quality i can get, but on DVD people have high expectations, and CVD isnt high enough, but i suppose if you want 3 films on one disc its an ok alternative.

Kwag even though i dont always agree with what you say, i dont agree with the post on ISO News, you have to learn to take things less personal, i dont use your templates because my current DVD player doesnt like Mpeg1 at anything other than the standard, and i can put |Mpeg2 720x576 8000kbps CBR with 384k sound and it plays perfect, so i can set my VBR rates very high, i just encoded Training Day ant it looked amazing at only 13mb a min for video only, on your 2 disc template it worked out at 10mb a min, didnt look as good but looked damn nice for the filesize.

What im trying to say is dont try and convert everyone to KVCD as its not to everyones tastes and dont take every negative comment, lies or not, as personal as it sounds like you have, i may be ranting, but i like this forum and i dont want it getting as bad as the ones on ISO News and VCD Quality.

[vhelp]

evening waldok,

just out of curiosity, what DVD movie were you trying to CVD to, and
what was the scene??

The reason I ask, is because I don't seem to notice those jaggies, and
I'd Sure'as HECK like to give the scene a go and see if I notice any
jaggies. I did a few scenes from the DVD movie (tv series) Farscape,
1-Nerve, 2-The Hidden Memory - I sure noone here has it though.
But, I didn't seem to notice it, and just did a few of the 2nd, opening
scene when they show what happened previously.

Also, are you viewing on your PC monitor??
Of course you will notice jaggies. The smaller your res. the more notice
able you'll see jaggies. But, if you play you CVDs view dvd player on your
TV, you should not see any jaggies.
So, don't pay no mind to PC monitor viewing.

Please get back to me (us) here on this, as I'm curious to hear your
response.

Thank you.

-vhelp

[kwag]

I dont really care about if its a valid DVD resolution as i would be ashamed to stick it on a DVD, people dont really expect much of VCD's SVCD's and i like surprising them at the quality i can get, but on DVD people have high expectations, and CVD isnt high enough, but i suppose if you want 3 films on one disc its an ok alternative.

Kwag even though i dont always agree with what you say, i dont agree with the post on ISO News, you have to learn to take things less personal, i dont use your templates because my current DVD player doesnt like Mpeg1 at anything other than the standard, and i can put |Mpeg2 720x576 8000kbps CBR with 384k sound and it plays perfect, so i can set my VBR rates very high, i just encoded Training Day ant it looked amazing at only 13mb a min for video only, on your 2 disc template it worked out at 10mb a min, didnt look as good but looked damn nice for the filesize.

What im trying to say is dont try and convert everyone to KVCD as its not to everyones tastes and dont take every negative comment, lies or not, as personal as it sounds like you have, i may be ranting, but i like this forum and i dont want it getting as bad as the ones on ISO News and VCD Quality.

But Martyn!, where have you been lately
We're not talking about MPEG-1 here. We're talking about MPEG-2
The SKVCD is MPEG-2, not MPEG-1!
And I'm not trying to convert anybody, I'm trying to let everyone see what they can get!
Did you actually see the SKVCD SAMPLE#1 ??
I don't think so, because you wouldn't have made this comments?
So why don't you download it, and then come back and comment on it
And talking about jagged lines, I don't see any jagged lines in this sample. Not even on my monitor, or my HDTV. That's part of the beauty of AviSynth's Bilinear resizing.

-kwag

[Martyn1980]

No Kwag your right i havn't been looking too much into SKVCD, even though i have the templates downloaded.
Just got the samples a minute ago, the reason i didn't really notice is i use your templates for Mpeg1 not Mpeg2, on the Yukai i have found the perfect settings for this player, and i dont need to change.

Your dealing with a template that will work for the masses im dealing with one individual DVD player, i use the Mpeg1 templates because of bad SVCD playback on the Toshiba, sorry though should have checked up before i posted.

As for the samples the farscape one isn't that great quality anyway looks like VHS. the text of the credits are particular bad, dont know the film of the second clip, but the jaggies are obvious on my TV, nothing spectacular but they are there, especially on Hugh Jackman's collar and Meg Ryan's hair at the end of the clip, and theres loads more that i can spot if i view it on the PC.

I hate jaggies even more than blocks and im baffled how i can see them bad on a 28" Widescreen and even a little on a 21" 4:3 TV, when HD people claim to not notice them, bear in mind i am incredible fussy though. and sit and watch Digital TV, looking for faults.

View the same clip at 720x480 and you'll see the difference, the image is too soft, and removes the clarity and sharpness of DVD that ive grow attached too, and no mass distributed template is going to give me the kind of results im getting right now, even though i did use CVD resolution to do a 2 disc Lord Of the Rings, which looked very good considering, but it was only a test and not the sort of quality i need to enjoy a film.

[adam]

It doesn't really suprise me. I think to a certain extent the dvd player might be causing some or most of the aliasing. The dvd player is what's doing the resizing. Everything has a tradeoff, 352x480/576 is a great alternative resolution for bitrates at or around 2mbits or lower, just expect to see aliasing and blurrier picture on some hardware setups.

I use both 352x480 and 480x480, depending on my source and my bitrate.