View Full Version : NTSC film PAL interlace reverse transform


yup
8th March 2004, 14:14
Hi folk!
Recently i buy strange DVD (National Geographic) MPEG-2 stream was bottom field first (Ii use bitrateviewer for identification) and interlace. I made deinterlacing use TomsMocomp ( prefer this filter for interlace material) video see very good beside landscape scaning scene (some jerke). Also during watch on DVD this scene see jerky. I try filter 24restore for this material, but result more artifacts. I think that material made from NTSC film 23.976fps to PAL interlace conversion. Advise me way restore source film material, if this possible.
With kind regards yup.

rfmmars
8th March 2004, 16:33
For a start I would go to www.100fps.com site. Many a time I need to improve a NTSC vcr recording where it doesn't have enough frames.

Bottom first is correct for NTSC.........What I did to slove my problem was to write a simple AVISYNTH script, I got all the information from that site.

richard
photorecall.net

Mug Funky
9th March 2004, 06:15
23.976 > PAL?

i suggest loading this stream in avisynth and separating fields.

if you see blends, then it's a converterbox job. if you don't it's possibly sped-up IVTC (i think NG use video and film interchangably, so it's probably converterboxed)

if there are blends, you're pretty much out of luck - there's currently no foolproof way of undoing this damage. best bet is a smart bob (there are a few around... search is your friend) and some kind of unblend plugin.

this will be slow and will still probably let some blends through though.

[edit]
by the way, a lot of the information on 100fps.com is a little inaccurate, but for the purposes of deinterlacing it's okay.

for example the colour interlacing it speaks of is an effect of colour sampling in the DV standard itself (all DV cams do this), rather than a hardware bug. if you were to deinterlace DV footage, it's vitally important you use deinterlacers that are aware of YV12 or 4:2:0 or 4:1:1 colour sampling (check the sticky "YV12 faq")

yup
10th March 2004, 14:00
rfmmars thank You for reply.
Mug Funky I try see fields with simple script
Mpeg2Source(...)
Viewfields()
some fields blended especially with with fast movie and scan.
Mug Funky may be this conversion film>puldown(NTSC interlace)>PAL interlace. I try restore 24, but work very slowly and I think very important starting point for restore progressive film or not?
with kind regards yup.

Mug Funky
11th March 2004, 12:27
blends... *shaking head sadly*

only way to get decent output is to smart-bob it (slow as hell) and then substitute the blended frames for it's adjacent frames (based on how "similar" they are).

this cannot currently be done flawlessly, and blends invariably get let through, but if you're willing to wait ~2 nights for an encode, then you'll get _very_ watchable results. (i did this for a backup of my box set of FLCL, as i'm always loaning the original to anime-ignorant friends and never get to see it myself :))

probably the nicest smart-bob at the moment is bobmatcher, but it's painfully slow. the search is your friend here... search for "smart bob" and take your pick :)

i have a smart bob script as well, faster than bobmatcher, but blurs the image slightly and can be over-conservative with sharp horizontals (it mistakes them for combing occasionally)

of course, after smart-bobbing you'll have to decimate to the desired frame rate. i don't recommend decomb's decimate, as it's job is to drop similar frames, so in the event that a blend makes it through the un-blender, it will _definitely_ make it into the final output, whereas with dumb changefps(x) you have a 50-50 chance of the blends not making it...

yup
15th March 2004, 17:57
Mug Funky thank You for fast reply!
I try use bobmacher script for deinterlacing and get strange result.

Case 1
Mpeg2source(?file.d2v?)
AssumeBFF()
Bobmacher()
Video looking jerky with backward motion

Case 2
Mpeg2source(?file.d2v?)
AssumeTFF()
Bobmacher()
Video looking jerky but better than in case 1
I use program BitrateViewer and Restream for identification field order, both program show BFF (I use DVDDecrypter for rip and DVD2AVIdg for make d2v file). And I want decimate 50fps to 24fps use SmartDecimate(),
but script string not work:
SmartDecimate(24,50,bob=bobmacher()).
On next day I will have enough time for experiment and wait advice.
With kind regards yup.

scharfis_brain
15th March 2004, 21:47
at first you should test with those ones:

1)
assumebff()
bob()


2)
assumetff()
bob()

that one, which gives the smoother result is the one to go for.

additional you can try some combination of this code-snipples
to change some weird kind of broken fieldorder:

a)
separatefields().selectevery(2,1,0).weave()

b)
separatefields().trim(1,0).selectevery(2,1,0).weave()

c)
doubleweave().selectodd