Star Wars Laserdisc Preservations. See 1st Post for Updates. — Page 4

g-force said:

Alright, I'm looking at the SW sample, and I don't see the crazy amount of temporal artifacts that the 720 mkv had, but it also looks less detailed than I remember the 480 mkv being (sorry, at work and can't compare). Anyway, the fine (good) grain seems to be gone and the big (some may argue ugly) grain seems to still be there. So, why bother? 

 

I've tried to take into account what you said about the detail and the grain.  When you have time would you look at this?  If you don't think it's an improvement I think I will concede with you and just release a 480p anamorphic version with only adjusted levels and saturation alongside the 480p raw letterbox.

It's the same method in the script above but I've added some masks and a couple other things to try and keep detail.  I might have cheated a bit, I'm not sure.  I'll wait to see what you say.

It's a split down the middle of the frame comparison of the raw capture on the right and the filtered version on the left.

Thanks so much for doing these, and sharing your work.
Also glad to hear you might tackle that Technidisc ANH in future.


Now some critique. I found these, what look like IVTC anomaly:


Frame 06457: Top field is GOUT 6457 (repeat) bottom field is GOUT 6458


Frame 41315: Top field is GOUT 41351, bottom field 41352


Frame 44432: Top field is GOUT 44468 (repeat) bottom field 44469

Thanks for the input and thanks for putting up the pics with the frame numbers. It will make it much easier to find and fix! I'll make sure and fix those tonight. You must have very keen eyes to find this kind of stuff. Is it possible to interpolate the last or first frame in a scene btw?

EDIT: ok they are fixed.  Could you also check in between frames 155980-155990 and let me know what you think that is?

The more I try to denoise it the more I have to agree.  I think I am going to just do an anamorphic conversion and let it be.  Should it be 480p or 720P?

Hi althor1138! I'm no expert, but I think a nice clean straight transfer of the laser disc is best. A what you see is what you get kind of thing. This I think, would be closest to an actual "preservation".


Thank you for the hard work my friend!

That was my intention in the beginning and then I thought I'd try to make it more presentable but like you guys say it just messes up detail etc etc. I think I've come full circle with it and I'm planning on releasing the 480p letterbox untouched and then an anamorphic that has levels adjusted and saturation adjusted if need be.

I know I'm teetering on the border of being a complete worm about the whole thing, but maybe anamorphic untouched? Please.

Forget that. I tried to edit my post and failed. 480p letterboxed untouched sounds sweet! Thanks again for your work on this one!

^you mean 960 nonsquare pixels stretched to fill a 1280 line?
I hope he won't do that ;-)

althor1138 said:

Is it possible to interpolate the last or first frame in a scene btw?

Yes, I'd definitely be interested in a list of frames to check.  The frames you are giving me match up exactly with my numbers. Is the Red Leader glitch only in my release?  It's the strangest one because it's only like half of the frame that is affected. 

I'll have to go back through the script to make it legible.  I've been cannibalizing it with all of these added fixes and I have a separate script for each side that feeds into one "master" avs file so I can kind of keep my sanity when sorting things out.  I counted the nnedi()'ed frames in my script and have now come up to 14 including the 2 from red leader.

Red leader glitch is also present in X9 capture and in Dark/Sega.
Pulldown cadence is unbroken, so that's my excuse for not finding it before now.
My eyes aren't as keen as you thought yesterday :)

The last frame (155989) comprise three fields, two of which are aligned, so you could probably fix it without using nnedi.
Adjust your pulldown() parameters [or selectEvery()] to pick the first two fields instead of the last two.
The earlier frame comprise two fields, one broken, so nnedi is probably the only choice there.

In the two captures I've worked with, it was possible using nnedi to recover the following GOUT frames:
52759
55748
55751
64443
70632
86682
92800
92803
97654
100687
102746
139231
158397


Here's from the X9 cap
That frame comprise three fields. Pick the first two

Double-posting to add more random thoughts.

Here we keep the top half of the broken field, and apply nnedi to the bottom half:
stackVertical( crop(0,0,0,240), crop(0,240,0,0).nnedi2() )
You could tweak the 240 of course. It seems to be different for each capture (Dark/Sega, X9, HF9G)


Is it just me, or these frames out of order?
You might just barely see left-to-right motion in Luke's ankle.
The flash frame, with glue at the bottom, should've been last.

I've tried two decoders. First DG's old, unsupported version
then tried newest FFMS2. Both times those frames are transposed.
Chew if you're reading this, what do you get from DG-nVidia app ?

I checked the GOUT and it seems these two frames are reversed.  If by some craziness in the script or on the laserdisc I'm not sure.

Weird, those frames are definitely out of order. Great find D_M! What did they smoke when they mastered this one, so many odd anomalies in it.

I'm going to wait a little bit before releasing this again because it seems we are finding little things to fix still.  When it seems that things are in order I'll put it up again.  Here is a little clip.  It is interleaved with 1 clip being the raw capture and the other being the adjusted clip I plan on using for the 720p version.  I've just adjusted the levels and tried to not crush or blow out any information.  I have applied a halo reducer which I think I'm going to use.  I wouldn't say it fixes the right-edge halo problem but I think it makes it better. 

althor1138 said:

I believe I also found the reason why some of the whites were getting clipped.  Mysycamore commented on this.  I'd have never noticed it otherwise. The 480p stream that I originally released was essentially the raw capture encoded to x264 but I did apply a levels tweak which I thought would not clip anything.  However, coring was by default set to true which automatically clips stuff when you use it(I think).  This time I have not adjusted anything whatsoever.  It's just been IVTC'ed and a couple wonky frames have been fixed but other than that it's untouched.  It looks a bit washed out because of this but is easily fixable with a few tweaks.  I plan on releasing it completely untouched this time however.

I skimmed through your 4:3 stream again and I actually don't know why I thought I noticed blown out whites the first time, there aren't any, I was talking crap so you may disregard what I said, sorry about that.

I was once again struck by how incredibly nice your cap is, that combfilter of yours sure knows how to deal with dotcrawl and rainbowing. :)

althor1138 said:

I'm going to wait a little bit before releasing this again because it seems we are finding little things to fix still.