IVTC and YV12

[ookzDVD]

@Forum,

Is there any working plugin to perform IVTC in the YV12 ?

[JohnMK]

For now, I believe DVD2AVI is all that's available. Correct me if I'm wrong, but there should be no problem using this as a frameserver in conjunction with YV12.

[ookzDVD]

Sure,

but the the source is "FILM at 95% or lower" as the DVD2AVI reported.
so it's not save to use the "ForceFILM" :(

Thank you.

[hakko504]

Try checking ForceFILM and deinterlace with TomsMoComp. Or wait until an IVTC plugin is available. neuron2 is reportedly working on an implementation of Decomb, but it could take a while before that is ready.

[ookzDVD]

@hakko504,

Thank you for your info,
I prefer to wait the Decomb ;)

[hakko504]

Wise choise. I'd do that too. :rolleyes:

[ookzDVD]

@hakko504,

Sorry for off topic,
do you have experiences using the trbarry's "TomsMoComp" filter
for "Pure Interlaced" material both for PAL & NTSC ?

I'm looking for the "good" parameters :)

Thank you.

[hakko504]

Well, I'm having somewhat mixed emotions about TMC. Since I prefer pure interpolation deinterlacing, I'm not overly satisfied with the ghosting that it creates. Still it seems to be the function that creates the least ghosting (of the functions that does blend the video), as well as the function that is easiest to compress. Recently I've had very good results with the SmoothDeinterlacer, the picture is almost perfectly clean when using that one, but it needs a lot more bits than TMC. It still seems like Greedy is the best compromise, giving a picture quality inbetween TMC and Smooth, as well as a bit consumption somewhere inbetween.
A quick comment about fielddeinterlace: This one seems to use more bits than Greedy and gives worse quality. Avoid.

As for parameters, I've always used the standard (1,15,1) or (1,15,0). trbarry pointed out in a thread here in the AviSynth forum that this is overkill, there is no point in going higher than 9 for the second parameter. First one is Top/Bottom, and the third is a verticalReduceBy2 that should be used if you aim for a low size.

[ookzDVD]

@hakko504,

Thank you for the answer,
I'll try to use the SmoothDeinterlacer then.

Thank you.

[trbarry]

As for parameters, I've always used the standard (1,15,1) or (1,15,0). trbarry pointed out in a thread here in the AviSynth forum that this is overkill, there is no point in going higher than 9 for the second parameter. First one is Top/Bottom, and the third is a verticalReduceBy2 that should be used if you aim for a low size.

I actually have backed off my opinion a bit more and now mostly recommend:

TomsMoComp(t,5,1)

where t is the TopFirst switch. Higher numbers for motion search will give a bit more detail but at the expense of a little speed and will also get a few more single pixel artifacts from false hits on the motion search. So to me SearchEffort=5 currently seems the best compromise.

For folks who want a pure interpolation from a single field you can just skip the motion search completely with:

TomsMoComp(t,0,1)

BTW, JohnAd on the DScaler team has just added a new algorithm to do the interpolation part of TomsMoComp in the DScaler version of it. I haven't tried it yet but it's getting good reviews from the DScaler testers so far. I'll probably add it to the Avisynth TomsMoComp also if it turns out to be an improvement (which sounds likely).

- Tom

[hakko504]

Thanks for the info Tom. A better interpolation together with the doublerate feature could very possibly be the thing that makes TMC the best deinterlacer.