A question about AQ in X264

[huang_ch]

In the "Blocky Backgrounds" thread, I've seen someone suggest to use AQ for some instances, though I've not met that kind of issue with my videos, I'm still interested in trying AQ, and I've found that AQ also bring an advantage to me which is that previously I have to force a much lower Quant for a special period of my video to get that period encoded to an acceptable target, but with AQ, I found that even I do not force the Quant, the quality is really improved a lot for that period, that's a good news to me.
But a bad news is that, after trying AQ, the overall bitrates increase by about 10%, and if when I higher crf from 24->25, I managed to get a similar bitrate than the original one, but comparing these 2 videos, the PSNR of video with AQ is much lower than the original one by more than 0.3, it's surprising. Can anyone give any explanation? Also is there any posts/articles explain how AQ patch works in x264?
Thanks.

[Audionut]

PSNR might be lower, but how does it look overall to your eyes.

And I think AQ might be more suited to 2-pass encodes.

[foxyshadis]

Months ago I got similar results as you when I tested a while back, but much stronger.

Among what I tested, overall visual quality at the same bitrate was always degraded. AQ requires significantly more bitrate to function well. But it'll also fix those occasional annoying blocky scenes where just pumping up the bitrate won't. My decision was I could live with a few bad scenes for the good of the whole.

It may well be that your material doesn't show the same kind of problems at all. What I tested against was mostly CG/animation.

I miss tcplx. ;_; I might go back to compiling it in myself and keeping another version around if I ever need lossless.