Quality Discussion HD/SD (Offtopic posts from Shana II Ep 1)

[Mentar]

Originally Posted by martino

Is this supposed to prove something, or be some kind of a deeply logical post...?

No, I'd say it's a silly trolling attempt. Note especially that he opted to choose a frame where the karaoke effect was a blurry-to-sharp transition. Unless of course PEDOS_GRANDE realized for the first time that these sources have poor frames too, and he felt the urge to share this insight. Especially since it was never in question that several sequences, most notably those taken from the first season, were normal SD to begin with.

But if our whining detractors feel the need to diss the source or the encode, let them. It's just wasted time

[PEDOS_GRANDE]

Originally Posted by Mentar

No, I'd say it's a silly trolling attempt. Note especially that he opted to choose a frame where the karaoke effect was a blurry-to-sharp transition. Unless of course PEDOS_GRANDE realized for the first time that these sources have poor frames too, and he felt the urge to share this insight. Especially since it was never in question that several sequences, most notably those taken from the first season, were normal SD to begin with.

But if our whining detractors feel the need to diss the source or the encode, let them. It's just wasted time

I dont think that frame was from season 1.

This one doesnt appear to be from s1 either

http://www.shanatan.org/mhd_still1.png

[Mentar]

Well, I already told you. It's all a sham, captured in 320x200 by handycam, and then vectored, upsized and repainted by trained monkeys ... and those ~15.000 people who download the MHD torrent each week are just tricked retards misled by the evul HD defiler propaganda bureau, and all those enthusiastic reviews were paid for.

Damn, you really caught me ^_^

Now how about you go elsewhere and do something useful for a change? Like, creating a fansub? Or maybe simply a better release purely in SD?

[DryFire]

Or finding a way to kill this thread...

[ArabianSwan]

wow.. who let Wild-Mentar run loose hehee
Anyway.. back to topic and bringing it to life with the following question:

Those shows stamped with "Hi-Vision" logo, were they true HD aired in an SD station?

I always thought, or still believe, Japanese use HV for HD materials, yet I saw this kind of complains when I did Sola. In both cases, yes or no, honestly I don't care. As long as I see a perfect 720p raw, I use it without second thought

[Orochimaru83]

Just because it's got HV logo doesn't mean it's really HDTV.
Most of the Nagasarete Airantou raws sporting that logo were really shitty and didn't look at all like HDTV.
While another raw capped at 1024x576 not spotting that logo did look like a true HDTV.
There's a lot of shows out there like that.

I checked all those 720p raws out there for Shana 2 (especially the 3-4 cappers who're well-known for doing awesome caps) and I agree with Mentar for it.

It was mastered in HD. Lots of details inside for a simple SD...

Let's enjoy Shana dudes

[Onniguru]

Originally Posted by TheFluff

bicubicresize(852,480).bicubicresize(1280,720)

Quick question that is unrelated to the current discussion...

AMV lists the proper resizing of an "NTSC anamorphic 720x480 footage" to be 848x480. The exact page is HERE. You list 852x480. Is AMV wrong? Was it a typo?

AMV is a pretty authoritave source for this, but I think YOU qualify as an even more authoritave source .

If A&E is wrong, I have some d2v-source avs scripts to fix...
(no not ones I distribute. I don't distribute DVD rips...)

[martino]

In this case you do not care about mod16, but go as close as possible to the resolution of the frame size with a 16:9 aspect ratio (which should be 853x480, but you've got the mod2 limitation there), since you are not passing this resolution into the DCT based encoder application.

EDIT: I guess the point is there, but someone with better technical knowledge should explain this better... I guess.

[Onniguru]

Nah, you answered it well enough . Thefluff was creating a png for viewing on these boards, so he kept the true aspect ratio as close as possible, in order to make his point. A&E is describing how to prep for encoding, so nat. they gave the nearest mod16 resolution .

[TheFluff]

Yes, that was the point. 852x480 is as close as you get to 16:9 while still being mod2, and it should be the resolution for anamorphic 16:9 encodes, while if you actually encode 16:9 with 480 pixels vertical resolution you should go either 848x480 or 864x480 to keep mod16'ness.

[Mentar]

For the sake of completeness: The best way of course would still be to go for anamorphic encoding and go for 720x480, or usually even better, 704x480 (with the sides cropped away). Since you're encoding in h264 in the first place, you can consider the ability of anamorphic playback for the viewer as basically certain.

[ForceDestroyer]

Originally Posted by Mentar

And to complete the circle of my argument: Does that mean that you can take any source and just use the same technique to create MHD out of a SD source? In my opinion, no - you can apply the same technique, but the quality gain depends on the detail grade of the source. If the source itself has been created HD, the MHD trick works, because there are enough details for the upscale to trigger, and to avoid making the upscale look like a mere oil painting. If the source itself is merely conventional SD quality, it does not.

I don't get why you keep feeding the trolls by redefining HD. It's irrelevant whether an SD source was created by HD downscale, a scanner, or MS Paint! The level of detail counts. And the maximum level of detail is probably higher when people aim for HD, thus your correlation. But it is definitely possible to get a better upscale from an SD source than from a HD->SD downscale due to resolution interference on downscaling. In reality, of course you are right - the HD downscales look better because the ones creating only for SD don't really care about image quality anyways.

As to all these image "tests" so many people did: To revert an upscale, you have to use the exact same method backwards the upscaler used. I think this is what most people in this thread didn't get. The exact method of the original upscaling the station did is very important if you try to proove anything by reproducing it. This is why results differ. You have an upscaled SD source. People who hit the mark by choosing the method to re-downscale it get better results than others.

Let me try to make rules out of this:

• The maximum possible quality is defined by the minimum resolution used at some point in the process. Any results below the maximum are meaningless, as screwing up is easy.
• Most image processing is NOT bijective and thus IRREVERSIBLE. Even if it theoretically is reversible, in reality it's usually not because of differing methods used.

So it's pretty simple really: If you get a decent image, do not mess with it if possible. If you get it messed up, try to carefully make it bearable again, damaging as little information as possible.

Eclipse did the RIGHT thing NOT resizing the image again. Blame the ones who created the original source if you don't like the quality!

one more thing: As far as I'm concerned, all quality losses discussed here are minor compared to the first HD->SD downscale. I must completely disagree with Mentar's comparison of HD->SD->HD scaling with HD->HD->HD encode. The difference in detail is massive, proving the theoretical numbers. See Gundam 00 or the likes. I've watched the opening so often now - it's just beautiful. Unlike any TV rip I've seen before. But probably it's just the first one I saw that was neither high-compression reencoded before the final encode nor ever resized to SD.