[x264] Should I encode interlaced or deinterlaced?

[dbmaxpayne]

Hi,

I'm sorry, if this has already been discussed (I'm pretty sure it has been :-) ), but I could not find something using the search.

I want to encode some old DV tapes with x264.
Now I questioned myself what would be more useful:

1. Using an avs-script with yadif to deinterlace the source and then encode it progressive

or

2. Directly encode it interlaced and let the decoding filter do the deinterlacing stuff.

What do you think is the best way?

Mark

[LoRd_MuldeR]

There is no "best" way. It totally depends on what you want to achieve ;)

I personally would always deinterlace before encoding, because I want to have full control of the deinterlacer and see the results beforehand.

Also encoding in "interlaced" mode (which is unavoidable, if you don't deinterlace!) compresses less efficient...

[Blue_MiSfit]

The "best" quality will usually be bobbing, and then encoding at double rate. This also tends to be the slowest (especially if you use a really "good" bobber), and requires the most bitrate.

Personally, if you're going for archival, I'd just encode the sources interlaced at CRF18 or so, and let my video card deinterlace them on playback.

Of course, you should do tests, and not listen to me ;)

~MiSfit

[nm]

Also encoding in "interlaced" mode (which is unavoidable, if you don't deinterlace!) compresses less efficient...

But it's more efficient than encoding bobbed (50p/60p) video. Personally, I'd bob my home videos to keep the full temporal resolution. For Blu-ray compliant output, you'd also need to resize the bobbed video to 1280x720. Or just encode as interlaced PAL/NTSC.

Using TGMC instead of yadif may be a good idea if you can spare the CPU cycles.

[dbmaxpayne]

dammit.. interlacing just sucks hard :-)

I've worked with some deinterlacers in the last few hours..
I noticed, that crippling the temporal resolution with deinterlacers like yadif and so on is very ugly (video stutters), so I will go for an bobbing deinterlacer, I think.

--crf 18 also gave only ~10% more bitrate for the bobbed encode (doubled framerate), so thats ok for retaining the full temporal resolution.

btw: --crf 18 --bff gave ~20% higher bitrate (maybe because weighted-p is not used when encoding interlaced)

But for TempGaussMC, which EdiMode (btw what is that?!?!) do you prefer?
NNEDI, EEDI2 or Yadif?
Currently I use TempGaussMC_beta1() without any parameters, because I don't understand them.

[audyovydeo]

I want to encode some old DV tapes with x264.

I have lots of DV material too.
As you may have gathered, there aren't any rules, except maybe : keep your DV.AVI files as the ultimate source, don't change into anything else for archival purposes.

what I do :

for shitty material : straight TDeint
for quality stuff : TDeint + NNEDI2


Function adeint(clip Input)
{
interp = nnedi2(Input, field=0)
# change field to 1 for TFF
TDeint(Input, mode=0, order=0, edeint=interp)
}

# test field order
# AssumeTFF().SeparateFields()


# TDeint(0,1,1)
adeint()


cheers
audyovydeo

[Blue_MiSfit]

Just FYI, YADIF can bob-deinterlace as well. Just use YADIF(Mode=1) (or YADIFMod for that matter :))

~MiSfit

[nm]

--crf 18 also gave only ~10% more bitrate for the bobbed encode (doubled framerate), so thats ok for retaining the full temporal resolution.

btw: --crf 18 --bff gave ~20% higher bitrate (maybe because weighted-p is not used when encoding interlaced)
Iteresting, I encode fairly detailed PAL sources (DVB broadcasts) at CRF 22-24, and bobbed video tends to require 20-30% more bitrate than encoding as interlaced, at same visual quality. Which bobber did you use for that test?

But for TempGaussMC, which EdiMode (btw what is that?!?!) do you prefer?
NNEDI, EEDI2 or Yadif?
EDI stands for Edge-Directed Interpolation and it's used for upscaling fields to full frames.

Advice from Didée on NNEDI2 vs EEDI2: http://forum.doom9.org/showthread.php?p=1374318#post1374318