Sign In

Help with interlaced video in After Effects — Page 2

Author
Time

Perhaps I should clarify how I am encoding the video for DVD.  I'm using TMPGE, which has a setting for the frame rate of the incoming video as:

"23.976 (internally 29.97)"

That tells TMPGE that although the video file format is labelled as 29.97, the actual frames in the video look correct if played at 23.976.  I also tell it to set the pulldown flag.  It then takes all that into account and produces an MPEG2 file that looks correct when played on a DVD player.

This is why I don't want the frame rate settings in the video altered if it can be avoided.

If I were really being picky, the frames look correct if played at 24fps.  But since I haven't had any problem with audio drift, I've just let TMPGE do its thing as described above.

"Close the blast doors!"
Puggo’s website | Rescuing Star Wars

Author
Time
 (Edited)

The source is 24fps, but what you gave Harmy has a frame-rate of 29.97fps.  All the original frames are there, but more than 24 of them are being shown every second.  To restore the original frame-rate, you have to slow the video down.

I don't know what you did with PG and PSB.  If I can find them, I'll have a look.

EDIT:  Sorry, Puggo.  I didn't see your second post.  The setting you used in your encoder probably explains why you didn't have to slow down PG and PSB

EDIT:  I found PSB.  Everything looks fine, so what you did obviously worked. :D

EDIT:  Harmy, that problematic clip looks just the same as the others to me, and the same script works.

Author
Time
 (Edited)

OK, I'm sorry guys, all this was pretty much for nothing - the combing in the problematic clip is actually already there in the source video - something must have gone wrong with Puggo's capture for that particular file. The whole issue was, that AE recognized the files as interlaced, so it set "Separate Fields" to "Lower Field First", which resulted in loss of vertical detail and therefore in both the jaggies and the hiding of the combing artifacts- once it was set to progressive, the jaggies disappeared and the combing became visible. I have another capture from Puggo of the same section, which however has badly blown out whites, so I will have to see what can be done.

Author
Time
 (Edited)

Actually, upon even closer inspection, all the files have those artifacts but they don't seem to be normal interlacing artifacts, because they show up as these short comb-like lines along the edges of vertical lines, and they are short, regardless of the speed of the movement and they don't appear in all frames. Here's an example - these are taken from three consecutive frames, the 1st of which doesn't have the artifacts and the other two do:

http://s25.postimg.org/tpxtir21b/Jabba.png

(I doubled the size to make the artifacts clearly visible.)

EDIT: In After Effects, thwere is an Effect called Reduce Interlace Flicker, which seems to take care of these artifacts without any significant loss of detail. I'll post a sample soon, for Puggo to decide, what he'd rather have.)

Author
Time
 (Edited)

OK, Puggo, here it is:

http://screenshotcomparison.com/comparison/88830

I would recommend the Mild Sharpen + Reduce Interlace Flicker, because it gets rid of the artifacts and otherwise it keeps the image almost the same (as seen it the lower left corner of the comparison).

Also, don't worry, now that I figured out the problem, I should be able to get you the files in exactly the same format as you gave them to me :-)

Author
Time
 (Edited)

They all look great.  I also can do mild sharpening and reduce interlace flicker on my end, and I prefer to examine their effect on an encoded NTSC DVD.  So I think my preference would be to have just the color correction, without the additional processing.  Unless you think that those artifacts are caused by the color correction, or a byproduct of getting in and out of AE.

I'd be interested in other opinions on this, too.  Chewtoie?

I forgot - did you get that small additional video of the desert scene where C3PO and R2D2 are approaching Jabba's palace?

Thanks everyone!  Especially to Harmy.  It is going to be fun getting this project back up again.  Next up - sound sync!!!  Ugh, the most tedious step.

"Close the blast doors!"
Puggo’s website | Rescuing Star Wars

Author
Time
 (Edited)

Yeah, I did get that clip but I didn't use it in the end because I thought what I had gave better results. I can do just the color if you want but I'd still recommend doing this kind of processing now, rather than on an already re-compressed source.

Author
Time

Harmy said:

Yeah, I did get that clip but I didn't use it in the end because I thought what I had gave better results. I can do just the color if you want but I'd still recommend doing this kind of processing now, rather than on an already re-compressed source.

Ok, as long as you think that the artifacts are something other than what normally occur in NTSC interlaced video, then go ahead and apply the processing.  Just keep the sharpening very mild.

Am looking forward to seeing the color-corrected footage!  Thanks again!

"Close the blast doors!"
Puggo’s website | Rescuing Star Wars

Author
Time
 (Edited)

The sharpening is only done to counteract the blurring effects of the remove int. flicker filter. And yes, these artifacts are definitely something weird and different from normal interlacing artifacts.

Author
Time
 (Edited)

As I understand it, each frame of the 16mm film has been captured by a camera that interlaces each frame, using the DV 4:1:1 format, with the individual frames physically advanced at a 30fps rate through the camera.

If each interlaced DV field is woven back together with its associate, each resulting progressive frame should have no combing unless the film frame moved between each field being captured by the camera.

DV 4:1:1 only captures 1/4 of the chroma in each scanline, so the pixel equivalent of chroma for each frame is only 180x480.  If the expansion of chroma is not performed properly, you can get what look like combing artifacts because of the huge horizontal chroma compression.

Ideally, one would want to use an expansion algorithm that is the exact inverse of the DV compression algorithm, but IIRC all you get is a range of DV decoders that do not necessarily match exactly and give varying results.

IIRC, there were some favoured DV decoders that did a better job than others and there were reviews on Doom9 (Cedocida being the best IIRC).

Avisynth has a very specific function created to properly expand DV 4:1:1 called Reinterpolate411().

I found Reinterpolate411() invaluable when capturing Star Wars laserdisc frames via DV and analogue capture card in an experiment to see which was better for fidelity to the original many years ago.

It is possible your processing of the DV requires a better DV decoder and/or initial processing via the Reinterpolate411() Avisynth function.

Since you seem to be using Avisynth for some part of the processing, I would recommend including Reinterpolate411() and using Cedocida decoder if possible to see if results can be improved, before resorting to interlace deflicker and sharpening processes to remove those "combing" artifacts.