Interlaced/Progressive report from DGIndex

[Backwoods]

www.flmpro.com/samples/odd.m2v

DGIndex is reporting this clip to be interlaced/progressive over and over.
Something basic I have been doing is FX_UnblendTC()
Telecide(1)
Decimate(5) which gets rid of the blending but leaves me with one combed frame every 3 or 4 frames. You can see it more in this sample: http://www.flmpro.com/samples/alice_test.avi

Is it possible the source was a PAL VHS dumped to NTSC DVD, the original anime was recorded at 14fps and blended or I just have no idea what I am doing when it comes to decombing?

[Guest]

Your second link should be alice_test.avi.

The mpeg appears to be frame-blended progressive. So why are you applying IVTC?

I'm not going to install all the obscure filters I need to duplicate your script. Post a clip with just FX_UnblendTC() applied.

The alternation of progressive/interlaced is a pathology of certain MPEG2 encoders.

[foxyshadis]

Looking at it, it's pretty much exactly what my function was made to reverse, with no bizarre PAL conversion throwing a monkey wrench in it. If mine alone gives poor results, try it with momonster's beta function near the end of the thread, which is a little better than mine.

On your segment both give good results. (You still need to use decimate after the function though; like telecide, it leaves 4 progressive and one dup frame. It's been brought up that even decimate may wonk it, see the end of the thread.) If you find a segment where one fails horribly, don't be shy to post it!

Also, what show is that? It looks familiar. :p

[scharfis_brain]

that mpg shows how to not deinterlace NTSC telecined stuff.
It is blend deinterlaced anime.
very tough to restore, cause there is double-blending.

[Guest]

It is blend deinterlaced anime. I wondered about that. But it's an mpeg not an avi, so I decided not to commit myself. :)

Usually you see residual combing when blend deinterlaced, but I didn't see any. You think they just did a dumb non-adaptive blend deinterlace? What gives that away for you, scharfi? I never see two blends in a row, which you'd expect if they deinterlaced pulled-down material.

[scharfis_brain]

hmm. that is strange.
But on the Pan-scene you can see the typical artifact of blend deinterlaceon 3:2 material. cause there are 3 clean and 2 blended frames, whereby the two blended frames 'hide' one frame, which only will be accessible by subtractive deblending starting with the clean neighbors.

[Guest]

hmm. that is strange.
But on the Pan-scene you can see the typical artifact of blend deinterlaceon 3:2 material. cause there are 3 clean and 2 blended frames, whereby the two blended frames 'hide' one frame, which only will be accessible by subtractive deblending starting with the clean neighbors. I think you're right. I only stepped through the beginning of the clip. Maybe that just has insufficient motion, such that the pattern doesn't show up (sort of the same reason that guide=1 sometimes doesn't work good on low base rate animation).

[foxyshadis]

hmm. that is strange.
But on the Pan-scene you can see the typical artifact of blend deinterlaceon 3:2 material. cause there are 3 clean and 2 blended frames, whereby the two blended frames 'hide' one frame, which only will be accessible by subtractive deblending starting with the clean neighbors.
That's what my function restores. ;) fx_unblendtc finds and uses both bad frames (with a bit of artifact protection for noisy encodes) to return two copies of the original frame.

It was probably never interlaced in the first place, just telecined and later blend-ivtc'd.

[Backwoods]

The alternation of progressive/interlaced is a pathology of certain MPEG2 encoders.

Understood.

Looking at it, it's pretty much exactly what my function was made to reverse, with no bizarre PAL conversion throwing a monkey wrench in it. If mine alone gives poor results, try it with momonster's beta function near the end of the thread, which is a little better than mine.

On your segment both give good results. (You still need to use decimate after the function though; like telecide, it leaves 4 progressive and one dup frame. It's been brought up that even decimate may wonk it, see the end of the thread.) If you find a segment where one fails horribly, don't be shy to post it!

Also, what show is that? It looks familiar. :p

MOmonster's latest version seems to do a great job. I'll have to encode the whole movie but from what I previewed it looks great.

Thanks alot for the help guys.


Also, what show is that? It looks familiar. :p

Monkey Punch's "Alice."