Is there a complete decimate function? bit depths, chroma subsampling ?

[poisondeathray]

Is there a more complete decimation function that supports other bit depths, chroma subsampling somewhere?

vapoursynth's vdecimate is limited (missing some modes and functions of tdecimate)

avisynth's tdecimate, decomb decimate work in 8bit YV12 or YUY2 only

Or is there a way to calc the metrics on a YV12/YUY2 clip and apply to another clip in the native pixel format?



Or I would like to request a tdecimate mod which supports other bit depths/sampling/pixel types

[Myrsloik]

Use clip2 in vdecimate. Done. Maybe I should add other formats too some day...

[poisondeathray]

Use clip2 in vdecimate. Done. Maybe I should add other formats too some day...

But there is no cycleR parameter in vdecimate, or other usage modes

For example, if you needed some other ratios, not just 1 in N decimation

[poisondeathray]

I guess a workaround for M in N decimate would be to run vdecimate a few times , with each iteration changing the value for cycle N-1

[Myrsloik]

I guess a workaround for M in N decimate would be to run vdecimate a few times , with each iteration changing the value for cycle N-1

That would be one naive way to do it, yes. One of the two main reasons I haven't bothered with M in N is that I've never really needed it myself and I don't have sample material to even try it on.

The other being that some people will want some kind of cyclic decimation with the dropped frames evenly spaced out, some will want most similar frames dropped and some want enormous cycles. It's almost easier to dump the difference and scenechange metrics from vdecimate and calculating which frames to drop from that then. Use a bit of python scripting and get the best results on your source!

[poisondeathray]

It's almost easier to dump the difference and scenechange metrics from vdecimate and calculating which frames to drop from that then. Use a bit of python scripting and get the best results on your source!

This is not so easy for python noobs (ie. me :D )

But I'm curious and willing to learn

Do you have any practical usage examples ?

[Boulder]

It's almost easier to dump the difference and scenechange metrics from vdecimate and calculating which frames to drop from that then.

Isn't that what mode 2 + 2-pass processing in TDecimate actually does?

[Myrsloik]

Isn't that what mode 2 + 2-pass processing in TDecimate actually does?

I guess it more or less does that. For some reason I've never really liked that 2 pass stuff.

[Boulder]

I guess it more or less does that. For some reason I've never really liked that 2 pass stuff.

I've found out that for example with MacGyver NTSC DVDs, it's the only way to get at least close to a non-stuttering output. The telecine pattern is a bit random probably due to the editing made back in the 80s. A normal IVTC looks mostly fine but when you notice the incorrect dupes or frame drops, you cannot unsee it anymore.