Better to go slightly larger or smaller?

[DRP]

When converting commonly available AVI files of say 512x384 kind of dimensions and wishing to convert them to (X)SVCD for playing in a DVD player, is it better quality wise to encode them in standard SVCD format 480x576 (talking PAL only here) or XSVCD 720x576 size?

I can't figure out whether to go larger or smaller when the horizontal size of the source fits in between the legal sizes of 480 & 720. I would appreciate any advice on what should produce better video quality at a given bitrate.

I've done a quick test encode of both and watched them to try and determine this for myself and my initial impression is that going larger with the 720/XSVCD option produces the better quality picture with fewer macroblocks but I can't explain why this should be the case technically speaking. I would have thought that as with JPEG images, enlarging reduces the resolution by making it look "blocky" whereas reducing the size can be done without resolution loss.

Any advice greatly appreciated.

[lordsmurf]

Image-wise, it is always better to shrink rather than enlarge.

[DRP]

Image-wise, it is always better to shrink rather than enlarge.

And does that transpose to video as well?

I suppose the comparison image-wise would be that you have a wallpaper of say 1024x768 resolution but your wallpaper displaying program is only capable of resizing of the fly images of either size 800x600 or 1280x960 resolution and your screen size is set to 1152x864.

Is it better to manually reduce the original image to 800x600 so that the wallpaper program can enlarge it to 1152x864 for display or would it be better to manually enlarge it to 1280x960 so the wallpaper program can reduce it to 1152x864?

This is really what I want the answer to if possible I think.

[lordsmurf]

If you do insist on a resize, stay with Lanszlos (or however you spell it) and not bi-cubic. But yes, it would be better to shrink rather than enlarge. Imaging theory stays the same, rather it be still or motion.

Put it this way: you can always REMOVE data, but you cannot add it.

Anything you add is fake, being copied from another area of the image.

[DRP]

I'm not wanting to manually resize my source AVI at all. The thing I'm getting at is that I want to view them on my TV using a DVD player. My DVD is capable of reading both 720x576 (XSVCD) sized and 480x576 (SVCD) sized MPEGs, and I want to use TMPGEnc to convert them.

So in this case, I put the unaltered AVI in as the source and then I have to manually set the output encode size as being either 480 or 720 according to my choice. I understand that you are recommending I choose 480 here unless the source of the AVI is actually 720 or higher (highly unlikely!), but surely, then the smaller 480 sized MPEG has to be stretched or enlarged somehow by the DVD player's internal software to make it fit the TV screen fully according to my preferences. I guess the 720 one would have to be stretched as well but obviously not by as much. In this case wouldn't it be better to have the DVD player doing as little resizing as possible? Surely the DVD player cannot resize on the fly with as little loss in quality as a computer can with all the time in th world?

Have I got the theory roughly right or am I talking sh*t?

[racer-x]

DRP wrote:

Have I got the theory roughly right or am I talking sh*t?

You're talking sh*t.

Look, lordsmurf answered your question twice, yet you don't believe him. Why don't you encode the video in both resolutions, put them on the same dvd, and see for yourself.

[lordsmurf]

Have I got the theory roughly right or am I talking sh*t?

You're got a somehwat handle on it, though still leaning a bit towards the sh*t. All of them are 4:3 aspect, so stretching doesn't really apply too much on playback. Stretching only applies when converting source A to source B on the computer, as you are manipulating pixels there. Plus 720x576 often has its own errors, not being a perfect 4:3 source ratio itself. And given that tv cannot often see beyond 300-500x576, more may be a waste.

I tend to lean towards 352x480 (or 480x480) for my encodes that come from 544x480 DirecTV. Only 720x480 when the source involved lots of editing with computer-created effects, which would be high resolutions.

Put it this way: how long is this thing? If it is just a few minutes, dump it on with 720x576 at 8.0 and call it good. If it's more, then your 720x576 will sacrifice bitrate and likely produce macroblocks if attempting keep the same time-length on the disc. A 352x576 or even 480x576 at 4.0 (or even 5.0 MB/s jus tot be safe) would be better than 720x576 at the 5.0 or 6.0 it would take to squeeze it on the disc. This only suggested since you're open to XSVCD. SVCD with higher bitrate.

At any rate, I'd say to try a minute or two both ways at a high-action or detailed scene, test on the CD-RW and see which one you like better.

[DRP]

DRP wrote:

Have I got the theory roughly right or am I talking sh*t?

You're talking sh*t.

Look, lordsmurf answered your question twice, yet you don't believe him. Why don't you encode the video in both resolutions, put them on the same dvd, and see for yourself.

Thank you for your helpful and insightful response here. In answer to a part of it, may I point you back at my OP where I said I have actually tried both different resolutions and to my eyes I found the larger of the two, ie. 720x576 to look the better with fewer visible macro-blocks.

This is why I am continuing to question Mr. Smurf's answer. I fully understand and accept the theory that he refers to in justifying reducing the size rather than enlarging, but to me, the end result doesn't support this theory and I don't understand why.

I think the problem here is that I don't understand what is going on inside a DVD player to get that MPEG on the TV screen. I can only compare it to displaying a slide show on your wall for example. If you take the projector and move it away from the wall, the picture gets bigger but it also starts to look grainy (disregarding focus of course), if you move it closer it gets much sharper at the expense of size.

If you have a small MPEG then for sure it's sharp and precise in image quality, but once it's up on the TV screen and filling it to capacity if it's a 4:3 ratio AVI it came from, then won't it get grainy (macro-blocks) as compared to a larger MPEG which doesn't need to be "enlarged" as much to fill the same sized TV screen?

What's wrong with this logic? If you know the answer, then please tell me.

[DRP]

Put it this way: how long is this thing?

It's 42:29 long - an episode of an NTSC TV show which I've converted to PAL.

If it is just a few minutes, dump it on with 720x576 at 8.0 and call it good. If it's more, then your 720x576 will sacrifice bitrate and likely produce macroblocks if attempting keep the same time-length on the disc. A 352x576 or even 480x576 at 4.0 (or even 5.0 MB/s jus tot be safe) would be better than 720x576 at the 5.0 or 6.0 it would take to squeeze it on the disc. This only suggested since you're open to XSVCD. SVCD with higher bitrate.

My DVD player cannot handle higher than 2520kbps to play, so what I've done as a test is use the Source Range feature in TMPGEnc to specify it to only encode the first 5 minutes of the AVI as a test. This I have done twice - once at 480 and once again at 720. Both times I used CBR @ 2520kbps - ie. the absolute maximum by DVD is capable of playing correctly. I then compared the look of those two different sized 5 minute long sample encodes. The 720 one looked better IMHO. The source AVI is sized at 512x384. In truth, I actually letterboxed the pictures inside both the respective 480/720x576 MPEGs to allow for the TV overscan, so the picture size was actually 448/672x544 but I don't think that's gonna make any difference.

[lordsmurf]

What qualified "higher quality" to you?

That's part of it too. What makes you think the 720x576 at 2520 is better than the 480x576 one? By all rights, a 720 encodes at anything below 4000k is dogfood.

If this was relatively still footage with a small color palette, and you prefer detail as the #1 qualifier in quality, I can see how the 720 may suit your tastes more, though it would still be by a marginal amount.

[FOO]

Note that resampling to a smaller resolution will
inherently apply a smoothing effect. With noisy video (sometimes)
that actually improves it IMO. You probably have a
good video and object to the loss of detail.

I have observed that with video that is basically 300-400 wide
(TV capture) and marginal bitrates, going to 720 makes it look worse.

[DRP]

What qualified "higher quality" to you?

It is true that the video I am using is not a high action piece at all. It is not a gunfight/car chase scene with high motion and frequent scene changes. It is a very slow moving, gentle scrolling/panning scene with head/shoulder shots of the characters talking to each other - definitely not high bitrate stuff.

The thing that convinced me of the difference in quality was areas of large block colour. In particular there is a shot of the sky at sunset looking over a lake which is slowly and smoothly camera panned across. During this one scene in particular there were what I can only describe as macroblocks in the large area of orange sky as the scene panned across in the 480 version. They were not there in the 720 version.

That's part of it too. What makes you think the 720x576 at 2520 is better than the 480x576 one? By all rights, a 720 encodes at anything below 4000k is dogfood.

If this was relatively still footage with a small color palette, and you prefer detail as the #1 qualifier in quality, I can see how the 720 may suit your tastes more, though it would still be by a marginal amount.

Relatively still? - Yes, definitely.

Small colour pallette? - Yes, definitely.

Do I use sharpness and detail as the main feature upon which I base my opinion on picture quality - Yes I most definitely do.

Is the difference I'm talking about only a very marginal one? - Yes it is only very slight.

I think I may have found my answer. Thanks for your time.

[DRP]

Note that resampling to a smaller resolution will
inherently apply a smoothing effect. With noisy video (sometimes)
that actually improves it IMO. You probably have a
good video and object to the loss of detail.

Yes. You're spot on here. The video is a very good quality (at least as far as I'm qualified to say so based on my relatively limited experience of working with XviD thus far) XviD stream and I am basing my opinion on the sharpness of the resulting picture. AFAIC sharpness is everything to me, both in picture and music. I like both to be crystal clear and precise. I view smoothing as dulling my overall perception and I probably translate that appearance as being of lower quality when in fact when viewed from a sensible distance away (as one normally would) it is not possible to pick the difference at all.

I have observed that with video that is basically 300-400 wide
(TV capture) and marginal bitrates, going to 720 makes it look worse.

Thanks for the advice. I was really looking for a kind of rule-of-thumb kind of approach I could use here and with this discussion I think I've found that. I think I will always use the closest DVD player legal size I can that is greater than the source video once I've taken into account the size reduction after adding black borders for the TV overscan etc. and then use the highest possible bitrate that I can obviously. I think this will suit me and what I watch best for the time being until I learn a little more on the subject and gain experience through doing more and more encodes.

Thanks for the help.

[Ejoc]

The 720 one looked better IMHO.

There's your answer, do what you think looks best.

I tend to lean towards 352x480

A CVD man