How much does video/audio encoder's efficiency heart/enhance input material?

[SealTooGreat]

With new codecs come new efficiency algorithms which claim "More quality in less space". But what is the price? - To have crippled material where half of non-existing-quality-things are reconstructed during decoder's postprocessing(pp).

It's your turn!

[foxyshadis]

Huh?

If you get more quality in less space by more accurately predicting what shoud have been in that space, what's the problem? You trade one set of artifacts for another set that are less distracting; switch codecs if you prefer something else's artifacts. People aren't using MPEG-4 over MPEG-1 just because it's newer, but no one's forcing you to.

Inloop and SBR are no more post-processing than residual reconstruction, anyway. They're an integral part of the decoding process.

[SealTooGreat]

foxyshadis, I don't have any problem about accurate prediction. And I've just wondered do words like "good" and "bad" efficiency exist or when you say "efficiency", does that always mean it's good.

Inloop and SBR are no more post-processing than residual reconstruction, anyway. They're an integral part of the decoding process.
What do some post-processings have to achieve in order to get full decoder's support if those stuffs ones used to be pp?

[Hellworm]

What do some post-processings have to achieve in order to get full decoder's support if those stuffs ones used to be pp?

Get into a standart and be a prerequisite for every decoder which wants to decode that standard, like inloop in h264 or sbr in aac.

And efficiency may be bad efficiency if something is of subjectiv lower quality even if it had a measured better quality. But the problem there lies in the definition of quality, for example ssim is a metric that complies much more to human perception than psnr. But all that is irrelevant because the pp that made it into the standards did increase quality both objective AND subjective.

[SealTooGreat]

So when pp is the part of standard is pp still pp or it gets another appellation? Or in another word does pp, once gotten into a standard, behave exactly like it was once when it wasn't part of standard.

[foxyshadis]

And efficiency may be bad efficiency if something is of subjectiv lower quality even if it had a measured better quality.

Metric efficiency is one thing, but subjective efficiency is the only one that matters for us humans. If one encode looks worse at some bitrate to you than something else, it's less efficient, regardless of what the PSNR tells you.

So when pp is the part of standard is pp still pp or it gets another appellation? Or in another word does pp, once gotten into a standard, behave exactly like it was once when it wasn't part of standard.

Being in the standard makes it more common. If deblocking had been part of ASP, standalone players and more software players might support it now, even if it was on the level of qpel or gmc, which didn't get widespread support for a while. Otherwise it's a vendor-specific extension, and even if five are doing it, the other hundred aren't.

Inloop is tied so tightly that disabling it when in use significantly reduces quality compared to not using it at all, unlike previous adaptive mpeg deblocking. SBR isn't as extreme, but it still sounds worse if the SBR part is ignored than if it had been encoded entirely in LC, since a non-negligible amount of bitrate goes to the SBR bookkeeping. (MP3Pro is the same way.) That's the difference that I'd consider between post-processing and advanced decoding.