Critique my AME CS6 settings: DV-avi ---> H.264 ---> WD TV Media Player

[newpball]

I've already talked at length about why I consider it a priority to keep the vertical dimension as is (discrete vs. continuous sampling).

It's not continuous anymore once digitized so, I am sorry, I must disagree with this argument.

[Cornucopia]

Disagree all you want, but the horizontal dimension was ALWAYS* an estimate, while the vertical was not. This affects many downstream processes (digital or not), including anti-aliasing.

Scott

*For analog originated material & interlaced material, which for DV & SD is 90% of these categories.

[LouGee]

Cornucopia, thanks, I must have missed some of your posts when I searched on this:
https://forum.videohelp.com/threads/365241-Is-720x480-DVD-source-conversion-to-720x540-upscaling?
https://forum.videohelp.com/threads/371211-Encoding-DVD-Movie-Does-the-DAR-Really-matte...80-vs-648x480?

Yes, now 640x480 makes sense.

Probably you're tired of questions like this, but if I could ask (on behalf of those working with DV): if you start with a dv-avi video that's 720x480 interlaced, and you're trying to convert it to square-pixel 59.97 progressive, why is it worse to upscale the vertical 480->540 instead of downscale the horizontal 720->640?

[Cornucopia]

First, you wouldn't/shouldn't convert to 59.97, but rather 59.94 (yours was probably a typo).

Next, there are "retained detail" benefits to downscaling, as opposed to upscaling, having to do with internal interpolation vs. extrapolation, as well as with oversampling.

Scott

[LouGee]

Thanks Scott, yes that was a typo - 59.94 it is. And OK, I always thought detail was thrown away when downscaling. I guess not. (and then throw on top all the other reasons to not change the vertical resolution, and it all makes sense)

AI: if you or anyone else is interested, I started a new thread about ripping semi-pro DVD's to eventually use with QTGMC for 59.94 progressive convertion:

https://forum.videohelp.com/threads/372016-Questions-about-Handbrake-DVD-rip-QTGMC-59-94-progressive?

[newpball]

why is it worse to upscale the vertical 480->540 instead of downscale the horizontal 720->640?

It is clear opinions differ on this matter, I'd say upscale to 720x640 that's 540 of course!

I suspect the downscale argument is mostly driven by "standards fetishism" rather than objective information theoretical arguments but I do not mind to be proven wrong ( "internal interpolation vs. extrapolation, as well as with oversampling" techno-lingo does not particularly impress me).

But to each his own, although unfortunately it seems that quite a few folks prefer to "resolve" difference of opinions using petty ad-hominem arguments. I suppose I should get my pill-box out

Edited to correct: 720x640 obviously must be 720x540

[Cornucopia]

High time!

*********

Upscaling to those settings would give one an improper display aspect ratio (1.125:1, aka 9:8).

Sounds like more sour grapes.

Scott

[newpball]

Upscaling to those settings would give one an improper display aspect ratio (1.125:1, aka 9:8).

Correct, I corrected the 640 with 540.

[_Al_]

The dv-avi file I've been using is exactly 60 sec (1 min) long, and has typically taken about 15min to convert... does this mean a 60min (1hr) dv-avi file would take about 15hrs to convert? (I haven't done one yet)

I just reliazed, with more DVavi files you can run more instances, for example if 3x ten CPU would be 100%.

For example drop couple of DV avi files (or much more) onto "make progressive mp4.BAT", and then do it again with different files, then CPU would be around 80%, ...., you do it again with some other DVavi files you need to re-encode and CPU would be 100%. So running more than 3 instances would make little sense, but if you want to force to work processors to 100% and using QTGMC without special Avisynth MT, it is possible, to go around and do that.

Doing this perhaps it would be a good idea to assign temp folder (in settings.BAT) to some other hardisk of yours or destination folder perhaps as well. Temp is using while encoding streams (video and audio). Destination is used while muxing those streams.

[Brad]

I suspect the downscale argument is mostly driven by "standards fetishism" rather than objective information theoretical arguments but I do not mind to be proven wrong ( "internal interpolation vs. extrapolation, as well as with oversampling" techno-lingo does not particularly impress me).

Here you posted a 1440x1080 zone plate scaled to 720x480 and then via 720x540 vs 640x480. I've highlighted portions of the section that you cropped. Very roughly eyeballed, orange is where the 720x540 becomes mushy, red is where the 640x480 breaks down, and green is where the 640x480 retains better definition than the 720x540.



The 640x480 looks mainly worse in the red section, but that's because this is an artificial scenario where the horizontal resolution has been constructed to be fully utilized. You should be able to show the same thing using something like Big Buck Bunny or other CG content. Neither case is the same as an image filtered by a lens hitting a sensor, that's then processed and compressed.

[_Al_]

And how about a samples with vertical lines?

[newpball]

And how about a samples with vertical lines?

This video has both horizontal and vertical components:



(Make sure you view in 1080p)