[Michael Morgenstern]

Thanks! I did some tests and confirmed, looks like it is after.

[Michael Morgenstern]

Thanks Darby! Even if there is a bit of a color shift, I know (hope) I can get rid of the greyscale shift… Also, as you can see the difference layer is colored so I’m hoping that will affect the cast.

I’ll check out your links! Question on one thing you said, though: “do essentially do what you’re doing now”

What am I doing now? I’ve researched Subtract, Difference, etc, and it’s not clear to me exactly what operations I should be doing. Subtract (F-A) -> Blur&Stretch -> Invert -> ?? to the original source?

Thank you!

[Michael Morgenstern]

Perhaps I explained this incompletely. What I want to do is take a small segment of the image, which doesn’t move, and getting the difference between that and the freeze frame. Then I will blur that difference, expand it horizontally so that it covers the whole image, and invert it / subtract it from the video footage.

Yes, the small section of the image this diff layer was taken from will be adjusted to be close to the image, but with the blur, the adjustment will not be perfect — it will just be a broad brighten/darken to even out the tones of the image. And the rest of the frame would be affected too.

So:
F = a pixel from the freeze frame
A = a pixel from the video
X = stretch(blur(F-A))

And I’m looking to get A+X
or A+stretch(blur(F-A))

[Michael Morgenstern]

I’ve tried to do that a few times and it hasn’t been great. The shift is so subtle that I don’t really have something to go off of, and since I have the adjustment layer in there already I’d much rather figure out how to make that one work.

I was also thinking of using some plugin that allows me to directly do arithmetic on the layers: for example, mapping the difference values to center around 0.5.

[Michael Morgenstern]

Alright, it’s been a few months and I’ve learned a ton on the subject since. This will be useful for you all to know, I think. I did the resize in FCP as suggested and the stair-steps were painfully evident. The thought of resizing an interlaced to interlaced frame was also just painful.
So I got Magic Bullet and used that software’s Instant HD package. It will do custom resizes as well as to HD, and uses the previous and successive frames to get a quality result. The resulting images have details in them that just didn’t exist beforehand.
Two downsides: one is the processing time. It took a full week of encoding, using two computers, to render my 46-minute documentary. I used frame mode which I would recommend, then you have to reconstruct the .mov Also, the plugin produces some bad frames with white flashes.
Take a look for yourself at the quality difference.
https://mjmfilms.com/comparison.zip
Cheers,
Michael

[Michael Morgenstern]

I did some tests by resizing some sample footage in Motion and Final Cut, with very surprising results. The unresized letterboxed video was the highest quality, and the Final Cut resize was very slightly degraded. The Motion resize, however, was very much lower quality, blurrier. It’s possible that settings can be tweaked on Motion to resize more sharply…

For both I created a DV anamorphic document and resized the footage to 133%. I exported with highest quality in both, turning on field rendering.

I’ve uploaded some layered Photoshop documents so that you can see them. Two have been deinterlaced in photoshop so you can directly compare the fields. The letterboxed original has been resized vertically in Photoshop for easy comparison. Take a look if you’re curious!

Frame 38;13 interlaced: https://mjmfilms.com/3813_interlaced.psd
Frame 38;13 deinterlaced: https://mjmfilms.com/3813_deinterlaced.psd
Frame 18;12 deinterlaced: https://mjmfilms.com/1812_deinterlaced.psd

Cheers,
Mike

[Michael Morgenstern]

David,
You’ve got my number – I am anal about this film. It’s going to festivals and I want to be sure that my shoddy camera work and shoddy equipment looks its best!
I found this interesting rant (https://www.dvinfo.net/conf/archive/index.php/t-137763.html) as well as https://discussions.apple.com/thread.jspa?messageID=7267937 , both of which appear to suggest that FCP is less than adequate for it. I’m being very, very stingy with my pixels since I have so few of them, especially now that I’ve chopped off so much of the frame, so I think I will go with Motion.
Cheers,
Michael

[Michael Morgenstern]

Wow, thanks everyone for all the help! Here’s my confusion: anamorphic DV has 480 scan lines. I’d be effectively mapping 390 lines to 480, meaning each scan line will be larger than one scan line in the new format, and it won’t sync up. The only way to make the odd scan lines sit on the odd scan lines and the even on the even would be to enlarge both fields individually, then reinterlace them. Follow?

Does FCP do this?
Again thank you, this is my first time on this forum but I think I’ll stop by here more often!
Michael

[Michael Morgenstern]

You’re saying that I could enlarge a 720×360 interlaced video to 720×480 without deinterlace?

[Michael Morgenstern]

The deinterlace has to happen to enlarge the film, since the source footage is interlaced. That’s why I’m concerned about enlarging it, because I lose some resolution on the deinterlace–even though the compressor can do a more efficient job.