Video quality comparission

[SAV]

I recently think out a method for numerical estimation
of video quality comparing to source material.

I have the source SOURCE.avs
(or SOURCE.avi, it is not very important)
I encode it to DEST1.AVI, DEST2.AVI, DEST3.AVI with
different codecs or different codec settings, it is unimportant.
The I write some .avs scripts like this:

cmp1.avs
subtract(avisource("SOURCE.avs"),avisource("DEST1.avi"))

cmp2.avs
subtract(avisource("SOURCE.avs"),avisource("DEST2.avi"))

cmp3.avs
subtract(avisource("SOURCE.avs"),avisource("DEST3.avi"))

(Also there can be added somithing like SelectRangeEvery(140,14)
to reduce processing time)

Then I make first pass (with high bitrate and quantizer=2 for DivX)
for each of this avs scripts.
(Nandub, XviD or DivX5, it is unimportant too, I think)

At last I compare estimated file sizes.
Most noise - largest size, less noise - lesser size).
What do you think about such comparision ?

[Latexxx]

What does the subtract command exactly do?

[evilhomer]

so does this actually correlate to quality?

[SirDavidGuy]

Originally posted by evilhomer
so does this actually correlate to quality?

It depends. How a video looks, and how close it is to the source are two very different things. Psychovisual Enhancements can make a source appear worse in this test, even when they look better.

[TheXung]

Yeah what he said. All in all, it's a pretty smart way to estimate quality without "quality measuring tools". which happen to also succomb to the same psycovisual deficiencies.