BadPAL-reconstruction?

[FreierGeist.]

Hi.
I think that this one, maybe, goes to Donald Graft...
I got a DVD NTSC source that was telecined from a PAL source and i used:
AssumeFPS(30)
Telecide()
Decimate(6)
crop(12,66,682,348)
BicubicResize(544,304,0,0.5)

The frame rate reconstruction seems to be ok
But it also seems that it needs a deinterlacer too (?)
Please, take a look @ http://hypokeimenon.hypermart.net/frame49465.jpg does that lines and columns ask for a deinterlacer?
Does it indicates something (bad telecine or bad ivtc)?
ps.: the screenshot was a VirtualDubMod with 2x zoom

thanks

[Guest]

You are running with postprocessing enabled, so this frame may have been combed and got deinterlaced by Telecide. The presence and perceptibility of residual combing depends on your threshold and the zoom.

Here's why. Your dthreshold is 9 for deinterlacing because you haven't changed the default. That means lines differing by 9 or less will not be considered combed and will be left as is. I have just done a little study of human contrast perception versus this dthreshold to try to determine the best default threshold.

One fact about human perception is that the larger the features are the less the difference that can be perceived. So by zooming in Vdub you get an unrealistic assessment of what is perceivable. I suggest assessing residual combing at the normal viewing size. Now, my study shows that 9 is a little high.

So try this. First turn on show=true and make sure this frame is in fact getting deinterlaced. It will be labeled as "Combed!" If so, set your dthreshold to 6 or 7 and view it at normal size. There should not be a perceivable problem. If the frame is not being deinterlaced, then lower threshold (different from dthreshold) until it is.

It may be that even at dthreshold 9 there is no perceivable problem at normal size. Remember that the lower you make dthreshold, the less benefit you get from the adaptive deinterlacing algorithm.

A good question is why this frame is combed in the first place if you have a progressive clip. To answer that I'd have to see a longer unprocessed source clip. If you want to post a link to it, I'd be happy to look at it.

[FreierGeist.]

Thank you so very much!
I'll do some tests and report the results!

[Guest]

I've received the VOB and had a look at it. First, this is not telecined, rather the rate was converted by field blending. That means you will not be able to use pattern guidance and you will not be able to decimate it without making it jerky. You should leave it at 29.97 fps if possible. If you must get it to 25fps, then go ahead but don't expect miracles.

Although you could just deinterlace this clip, there are enough clean progressive frames that using Telecide is a benefit. Therefore, following is the script that worked well for me. I also cropped it and did a little tweaking and sharpening.

I set the thresholds rather low because you were worried about residual combing. But you were zooming to look for residual combing and that is known to be misleading because our contrast perception improves as the feature size gets larger. Still, you may be planning to view this zoomed to full screen, so I made the thresholds low enough that it looked good that way.

Finally, note how I did the sharpening. This avoids sharpening any residual combing and making it more visible.

EDIT: I added Decimate(6) before the Tweak() call and it looked fine. I couldn't see any jerkiness playing it at normal speed.


loadplugin("d:\avisynthplugins\mpeg2dec3.dll")
loadplugin("d:\don\programming\c++\avisynth\decomb410b1\release\decomb.dll")
loadplugin("d:\don\programming\c++\avisynth\msharpen\release\msharpen.dll")
mpeg2source("edited.d2v") ## made with NON FORCE FILM
Crop(12,64,-28,-64)
SeparateFields()
MSharpen(threshold=10,strength=80)
Weave()
Telecide(threshold=10,dthreshold=6)
Tweak(bright=20,sat=1.3)

[N_F]

@neuron2 - Could you elaborate a bit on why you use msharpen on the separate fields instead of the frame? Or was it perhaps unique to this particular clip?

[Guest]

Originally posted by N_F
@neuron2 - Could you elaborate a bit on why you use msharpen on the separate fields instead of the frame? Or was it perhaps unique to this particular clip? This clip is effectively interlaced due to the high number of blends and lack of telecining. With any interlaced material, even after deinterlacing (with an adaptive filter) there will be residual combing. It just won't be perceptible. If you sharpen after deinterlacing, you may sharpen the residual combing detail and make it perceptible. The danger is somewhat less with MSharpen because you can set its threshold so as to hopefully avoid this. But as a general policy it is safer to do things like this.

If you deinterlace by discarding a field, this precaution is not necessary because there is no residual combing.

[FreierGeist.]

neuron2, thank you very much!

[N_F]

Thanks for your reply.