MeGUI: Clever (TM) vs. Resize

[Benihana]

Hello. I apologize for something that must be so simple to most of you, but I've become stumped just recently due to my lack of understanding. I have been trying to read as many articles/threads that I can find to help me understand this better, but keep getting stumped.

Background:
I have used GKnot extensively with XviD and mp3 for my video conversion needs. I have done thorough testing and have come to a workflow that I utilize to maximize my perceived quality to my liking. However, now that I'm trying to move on to MeGUI, x264 and AAC, I'm finding myself a bit stumped. I like to understand all the options available for my encoding, simply because I like to think that I maximize quality to my standards. Using default settings will work, but I'm always left wondering "well... what does this option do? Maybe it'll help compress the picture better with no noticeable loss of quality? I don't know."

Problem:
I cannot figure out the difference between using:
1) Clever (TM) anamorphic encoding: Resize to mod16
2) Resize: Suggest Resolution (mod16)

Source: Matrix, The [DVD]

What Happens:
- First, I use the Auto-crop to remove black bars.
- When I use the "Clever" option, I get a resulting video that is 720x352.
- When I use the "Resize" option with "Suggest Resolution" checked, I get a resulting video that is 720x288.

If I understand correctly, the "Resize" option keeps your SAR the same (as source), but will modify the DAR by finding a vertical resolution that is mod16. The "Clever" option keeps your SAR and DAR the same (as source), and merely finds a vertical that is mod16. Is this correct?

I'm confused because the "Clever" version in "AviSynth script creator" looks like the original source, however the "Resize" version looks squashed. My concern is two-part:

1) I'd like the resulting video to look the same (i.e. not squashed) when played back.
2) I think that I'd like to preserve as many of the original pixels as is possible, minimizing loss of pixels due to resizing.

I thank you in advance for helping an older encoder try to find his way again. :)

[smok3]

from http://mewiki.project357.com/wiki/MeGUI/Tools/AviSynth_Script_Creator Clever (TM) Anamorphic Encoding: Checking this box will enable anamorphic encoding.

What that basically means is that x264 will get proper signaling flag and you will keep most of your pixels.

[Benihana]

Ah, thank you for the reply. Yeah I had read that on the wiki project page, but still could not make sense of when I should use Clever vs Resize. Is my understanding of how Clever works vs. Resize correct?

If I understand correctly, the "Resize" option keeps your SAR the same (as source), but will modify the DAR by finding a vertical resolution that is mod16. The "Clever" option keeps your SAR and DAR the same (as source), and merely finds a vertical that is mod16. Is this correct?

[Benihana]

Quick question, when would it be better to use "Resize: Suggest Resolution (mod16)" over "Clever (TM) anamorphic encoding: Resize to mod16"?

[hello_hello]

Ah, thank you for the reply. Yeah I had read that on the wiki project page, but still could not make sense of when I should use Clever vs Resize. Is my understanding of how Clever works vs. Resize correct?

Almost.
Normal (non-anamorphic) resizing works as it always has. You're encoding using square pixels. You crop, pick a width and the "suggest mod16" function picks the closest mod16 height.
Using that method it's traditional to resize down to square pixel dimensions. Because DVDs don't use square pixels an anamorphic DVD may be (for example) equivalent to 1024x480 in square pixel dimensions, but instead of increasing the width to 1024 you'd keep it at 720 and reduce the height instead, so you might end up with 720x400. Is that what you're referring to when you say the resized version looks squashed? Basically it is, and you'll lose some detail that way.

If you think of the picture aspect ratio as the shape of objects in the video then in theory you'd want it to always remain the same. Because the above method resizes to the nearest mod16 dimensions then depending on cropping there may be some picture aspect ratio distortion. Some programs calculate how much distortion as you crop so you can adjust it accordingly. MeGUI doesn't.

The idea of anamorphic encoding is to use the same shape (non square) pixels as the original DVD (or close to). The "resize to mod16" anamorphic option is, I guess, a compromise between using the same shaped pixels as the DVD and sticking to mod16 dimensions. It should always keep the picture aspect ratio intact (as do the other anamorphic methods). It changes the SAR to the nearest mod16 dimensions after cropping (both width and height). The DAR also changes according to your cropping to keep the picture aspect ratio the same as the source, therefore the pixel aspect ratio must also change so you're no longer encoding "pixel for pixel". You're resizing a bit.
The rest of the anamorphic encoding methods involve no resizing. It's fine to encode using the "non mod16" method if you wish as it doesn't matter when encoding using x264. If you want to encode mod16 and don't want to resize at all, then "overcrop to achieve mod16" is probably the way to go.
Personally if it's only a matter of a few cropped pixels I generally use the "resize to mod16" method to keep the video 16:9 as I've compared a few of those encodes with non-mod16 encodes (no resizing) and I can't see a difference. Otherwise I encode non-mod16.

MeGUI also has an "acceptable aspect ratio distortion" option which is 2% by default. I assume to allow it to easily give you nice 16:9, 4:3 type aspect ratios when using the "resize to mod16" anamorphic method. I generally change it to 0% and check the script tab as I adjust the cropping until I've got the desired aspect ratio (which I generally only fuss over when it's 4:3).

I also never use the ITU resize method, only the straight 16:9 or 4:3 resize methods. I'm sure 99% of DVDs don't use the ITU resize method but I'm in PAL land. Maybe it's different in NTSC land.

Anamorphic encoding is the way to go. It'll retain a lot more detail than the "Resize: Suggest Resolution (mod16)" method. Some people resize up and encode using square pixels (1024x480 for example) but I don't see the point. Might as well use the same shape pixels as the source (or close to if you use the resize to mod16 method). Plus I don't think MeGUI will let you resize up that way anyway.

I hope the above makes sense... and helps.

[hello_hello]

Quick question, when would it be better to use "Resize: Suggest Resolution (mod16)" over "Clever (TM) anamorphic encoding: Resize to mod16"?

Almost never. Only when encoding for playback on a device which thinks all pixels are square. I don't think it's much of an issue any more, not like the old XviD/AVI days.