Can you help me with field blending (film to pal)?

[qwerty1983]

This is my small (12 sec) sample https://www.mediafire.com/?rb7b0ept1bltg39

dvb-t recording, h264, film to pal

Forget about the wrong aspect ratio (i'll fix that). There's field blending in this material. Is there something i can do? The least bad solution?

Thank you in advance

[Overdrive80]

If you use separatefields, will can see order fields wrong. Its possible that before you should fix it.

[Music Fan]

When I play it with MPC-HC or VLC, there is no problem. But if I open it with avisynth (LWLibavVideoSource), I see the problem : tfm() solves it.
Therefore, if you don't absolutely need to re-encode this file, play it with MPC-HC or VLC or a standalone player that should handle it correctly.
For the aspect ratio problem, you don't have to re-encode, just change the PAR (with Yamb, MKVmerge, TSMuxer ...).

[qwerty1983]

I used tfm(order=1,field=1) but eventually i think tfm(order=1,field=0) produces better results in this sample.

There's a problem. I use this script
LoadPlugin("C:\Program Files (x86)\AviSynth 2.5\plugins\DGAVCDecode.dll")
LoadPlugin("C:\Program Files (x86)\AviSynth 2.5\plugins\NicAudio.dll")
V=AVCSource("sample.dga")
A=NicMPG123Source("sample.mpa",normalize=false)
AudioDub(V,A)
AssumetFF()
tfm(order=1,field=0,mode=5,pp=7)


and i get these blocks in some frames (look at the picture). How can i get rid of them? They are not present in the original clip.
http://img244.imagevenue.com/loc340/th_369605116_sample_122_340lo.JPG (http://img244.imagevenue.com/img.php?image=369605116_sample_122_340lo.JPG)



Also, does it make any sense to use in the end qtgmc(inputtype=3) ? I tried that and i like the result.

[feisty2]

pp=7 is buggy
use pp=2 and clip2 function

[qwerty1983]

pp=7 is buggy
use pp=2 and clip2 function

i used tfm(order=1,field=0,mode=5,pp=2,clip2=qtgmc(selecteven))
but the blocks are still there. (even without using tfm)

[Music Fan]

I used tfm(order=1,field=1) but eventually i think tfm(order=1,field=0) produces better results in this sample.
What difference do you see compared to a simple tfm() ?

i get these blocks in some frames (look at the picture). How can i get rid of them? They are not present in the original clip.
Probably a problem with your source filter, I don't have these blocks with LWLibavVideoSource. You can also try FFVideoSource.

[qwerty1983]

What difference do you see compared to a simple tfm() ?
There is less blending
For example in field 369 look at the picture [tfm()-top and tfm(order=1,field=0)-bottom]. And it's the same for the rest of the fields. This is with AVCsource

http://img180.imagevenue.com/loc795/th_366136969_sample_field369_122_795lo.JPG (http://img180.imagevenue.com/img.php?image=366136969_sample_field369_122_795lo.JPG)



Probably a problem with your source filter, I don't have these blocks with LWLibavVideoSource. You can also try FFVideoSource.

Huge problems with FFVideoSource.

With LWLibavVideoSource there are no blocks but in some parts of the clip there is back and forth movement (like wrong field order on purely interlaced material)

I have the same back and forth problem with a few frames when i deinterlace (avcsource and qtgmc) purely interlaced material. In that case i firstly create an intermediate file (using a lossless codec) and then i apply the qtgmc filter.

[johnmeyer]

That clip has already been heavily processed by some sort of motion estimation technology. I looked at it using separatefields(), and it is actually quite fascinating: there is motion from one field to the next happening on the background of the scene, but the actor in the foreground shows virtually no movement between matching fields, even when he is moving. It is almost as though the background is interlaced (through some artificial means, of course), and the foreground is progressive.

I looked for signs of field reversal, since someone said they saw that, but I didn't see it on my setup.

SRestore might be useful, but I can't say for sure. I don't think I've ever seen a clip quite like this one.

[edit]Here is an animated GIF, showing the two fields from the same frame (I doubled each scan line to make it easier to view the result). The actor's hands are moving rapidly from his waist to his shoulders, but you don't see any movement between fields. The camera is panning from the right to the left, and you DO see movement between fields. This movement between fields from the same frame obviously was not in the original film. My conclusion, stated above, is that some sort of motion estimation software created this result. I don't see how any dumb field blend could do this.

http://i177.photobucket.com/albums/w208/johnmeyer/GIF%20Animation_zpsdsvh2tv7.gif

[qwerty1983]

I'm a little bit frustrated. Different source, different results and different problems. Look at this picture (it is field 369, i use separatefields)


http://img276.imagevenue.com/loc131/th_372710520_sample_field369_new_122_131lo.JPG (http://img276.imagevenue.com/img.php?image=372710520_sample_field369_new_122_131lo.JPG)



That clip has already been heavily processed by some sort of motion estimation technology.
All the movies from this era (50's, 60's) have been telecined this way from this company.

[qwerty1983]

Ok, i found a source filter that works with me

loadplugin("C:\Program Files (x86)\AviSynth 2.5\plugins\avss.dll")
dss2("sample.ts",fps=25)