Resolution & perceived quality: anamorphic or 1:1 PAR?

[SpikeSpiegel]

Supposing that you want to encode an anamorphic 16:9 PAL DVD with XviD using a specific "compressibility value", which resolution would you use:
720x404 (16:9) or 602x482(anamorphic 16:9)? (please, don't mind if they aren't mod-16)

In both cases, with the same bitrate & frame rate, BPF (Bitrate per Pixel per Frame) has the same value (for 720*404=290880 and 602*482=290164), then the compressibility should be almost the same, but...which may appear better?

Instinctively, I would say that, if a TV is used to display the video, "vertical resolution" is more important (then I would use 602x482), but I'm not very sure, then wanna know what do you think about it.

[Sagittaire]

For HVS vertical resolution is more important than horizontal resolution: for example 480*480 PAR 16:9 could be better than 640*360 PAR 1:1 ...

[lamer_de]

Dosn't the 602*482 get scaled to 720*404 during playback? So, I would imagine they do look the same, unless either the resizer during playback or the resizer used to bring the anamorphic sources to 720*404 is inferior to the other. While there probably is an advantage in higher vertical resolution, anamorphic encoding was invented to make better use of the already existing resolution conventions of the TV norms, and therefore really getting a higher resolution when using anamorphic if compared to letterboxed encodes (at least that's how i always understood it :p). As Spike doesn't need to follow defined width/height conventions, I doubt it makes a big difference if the output resolution is the same. But that's not proven by me in praxis, so I guess trying is the only way to find out ;)

CU,
lamer_de

[SpikeSpiegel]

Thanks for your quick answers!

Actually, using anamorphic resolution, the player's resizing method may nullify the vertical resolution gain:
IIRC, on standalone players, it should work enlarging the width (4example, scaling 720x576 [anam. 16:9] to 1024x576), but I fear that software players' one scales resolution reducing the height (maybe it's a coincidence, but, with MediaPlayerClassic in windowed mode, 4example, 720x424 [PAL 16:9 anam.] clips are displayed with the same size of 720x298 ones).

Yesterday I've done a quick test with my standalone player, but it's quite hard to do an objective comparison... with a 42" hi-definition LCD it should be more easy, but atm I don't have one :D

[SeeMoreDigital]

Hi Spike,

Quite a while ago I reported my own observations along these lines...

Most of my Mpeg4 back-ups are generated at 1:1 (ie: anamorphic NTSC 720x480 or PAL 720x576) for playback in hardware. But I decided to see what would happen if I converted the same sources to NTSC 848/864x480 (853.333x480 is not possible) or PAL 1024x576, at the same bitrate.... And I have to admit I was pleasantly surprised at the results.

Unfortunately, hardware playback of PAL 1024x576 Mpeg4 encodes is not possible (yet). Limitations in current chip-set design means they max out at 414,720 pixels. But interestingly, it is possible for me to generate and playback NTSC 848/864x480 encodes in hardware.

If you want to learn more, check out some posts by Soulhunter (some using custom matrices). He really took this activity to the next level and provided many images and created much debate...


Cheers

[SpikeSpiegel]

Originally posted by SeeMoreDigital
Hi Spike,

Quite a while ago I reported my own observations along these lines...

Most of my Mpeg4 back-ups are generated at 1:1 (ie: anamorphic NTSC 720x480 or PAL 720x576) for playback in hardware. But I decided to see what would happen if I converted the same sources to NTSC 848/864x480 (853.333x480 is not possible) or PAL 1024x576, at the same bitrate.... And I have to admit I was pleasantly surprised at the results.

Unfortunately, hardware playback of PAL 1024x576 Mpeg4 encodes is not possible (yet). Limitations in current chip-set design means they max out at 414,720 pixels. But interestingly, it is possible for me to generate and playback NTSC 848/864x480 encodes in hardware.

If you want to learn more, check out some posts by Soulhunter (some using custom matrices). He really took this activity to the next level and provided many images and created much debate...


Cheers

Hi!

Are you referring to the LimitedSharpen-SharpResize test?
If it's so, well, I agree with the fact that this method (upsize with avs filters) gives better results than realtime-upsize (used by software and "common" standalone players), but the problem here is a little different:

considering how realtime resizers (software or not) work and which device (CRT/LCD-monitor/TV, etc.) is used to display the video, is there any effective quality gain of using anamorphic PAR to increase vertical resolution without reducing compressibility (read: width x height product) ?

EDIT: Maybe you don't have to wait too much for 1024x576 hardware compatibility: I've red somewhere that SigmaII based new Kiss players will handle high resolution and bitrates!

[SeeMoreDigital]

There are limits to anything...

There are some people on the forum who seem to think re-sizing DVD 720x480/576 resolutions to 1280x720 gives them HD quality... This is simply not true!

In much the same way an 2.35:1 source, cropped and resized to say, 640x272 or 720x304 will not look good on a HD 42" plasma (at whatever bit-rate you use to generate your encode).... Don't get me wrong, such encodes wont look bad, but whenever you elect to fart about with the vertical resolution, there will be a noticeable loss in quality you'll never be able to recover fully. Regardless of the third party tools you elect to use.

The use of real-time re-sizers can be a contentious one, as we all possess the ability to react to an moving image differently. For-instance, I can't stand 100Hz digital processing (as used in many European wide-screen TV's) but I can put up with Philips' "PixelPlus" technology... both techniques can be used to enlarge an image.

I did quite like my old analogue "line-doubling" machine though!


Cheers