QTGMC Alternatives? (double rate deinterlacing)

[Stereodude]

What other high quality double rate deinterlacers are people using beside QTGMC for pure video content? QTGMC isn't working well for me on a particular clip due to it's inherent noise reduction characteristics (EZdenoise is off). The output getting blocky / banded in low detail "flat" areas with subtle gradient like shifts and I'm looking for alternatives. I'm trying out TDeint and Yadif. What else is out there?

If it matters I'm working with 1080i60 MPEG-2 footage.

[kuchikirukia]

Is the problem QTGMC or is your encoder just not giving those areas enough love? Try messing with AQ and qcomp.

[Stereodude]

Is the problem QTGMC or is your encoder just not giving those areas enough love? Try messing with AQ and qcomp.
After more testing it was mostly the encoder. Using placebo instead of very slow and some --aq-strength adjustments did the trick. For whatever reason --aq-strength works very differently with placebo than it does with very slow when crf encoding. :confused:

[kuchikirukia]

Placebo enables nofastpskip. I've read fast P skip can have problems in flat areas.
Nothing else it does strikes me as though it would have an impact. It's called Placebo for a reason.

[Stereodude]

Placebo enables nofastpskip. I've read fast P skip can have problems in flat areas.
Nothing else it does strikes me as though it would have an impact. It's called Placebo for a reason.
Yeah, I'm not sure exactly what's going on. I was working on a 720p60 clip and increasing --aq strength with preset veryslow and CRF of 17 caused the bitrate to drop as aq strength was increased and the flat areas did not improve. However with placebo the bitrate remained nearly constant as aq strength was increased and the flat areas improved visually as aq strength was increased.

However, when I tried the same thing on a 720p24 clip with veryslow the bitrate increased with aq strength and the flat areas improved. The same behavior occurred with placebo on the 720p24 clip.

Right now I'm a bit puzzled why aq strength + veryslow had a different impact on the 720p60 clip than the 720p24 one. :confused:

I'm going to run a bunch of compression jobs while I'm at work today as a sanity check and to gather data.

[kuchikirukia]

Unrelated to the your flat area thing, try qcomp 0.7. I'm running AQ mode 2 strength 1.0 on a 720p60 concert (haven't messed with that other than to notice lowering it destroys the picture), and raising qcomp to 0.7 seems to really wake up the picture. It raises the bitrate decently (I'm eyeballing 25% from the projected filesize) and I don't know if I'd be seeing the same effect if I lowered CRF to get the same overall bitrate, but wow. I just went from, "I guess QTGMC is a heaaavy denoiser," to, "Now THIS is a Bluray!"
e: (this is at CRF 18)


e2: Never mind. Adjusting for bitrate, there's little difference. What I was seeing was just that I needed that 25% more bitrate.

[Stereodude]

So my tests are partway done and the behavior is totally different from before. I don't understand how that's possible. I'm using the same base source and the same version of x264 and the same command line options. The only possible difference is where in the x264 command line I placed the --aq strength option. I guess I will try testing that possibility too.

*sigh*

[Stereodude]

Well, I give up... I tried everything to duplicate what I saw before with x264 and can't. I use the identical source file and the same x264 options and I'm just not seeing what I did before.