You_Too said:

It's quite complicated actually. Not like our GOUT project but still.

I did it all in avisynth. Since there was a lot of blue and cyan in the 97 SE I first used selectivecolour to lower those colors a bit. Then I made histogram files for colourlike, using some shots from the 97 SE and some from theatrical prints like the Senator screening.

I then made a 50/50 of the histogram conversion result and the cyan/blue fix so I got the best of both and finally, I made an additional RGB curve to correct the pink tint in the Mos Eisley scenes.

I admire your hard work! But it's too much for me... my restoration experience is limited, and my time too, as I have family, a dayjob, and... cats! ,,,^..^,,, and I'd like to release this project ASAP.

But some color correction is needed, so I tested a lot, and find a simple, but effective, method; merge the original laserdisc image (that is less pinkish than GKAR. by the way) with the same image, auto color corrected. I know it's not the perfect, definitive solution, just only a compromise - but a working compromise is better than pinkish images! (^^,)

Well, let me explains some of my later tests: I captured the french laserdisc (side 1 for the moment) with three different players, directly without passing through any dvd recorder; the players used are all Pioneer, CLD-D925 via S-Video, LD-V4300D and CLD-600 via composite; then I averaged/median the three captures, to obtain less noise; then I averaged/median the resulted image with the auto color corrected one. For comparison, I used GKAR, unprocessed, only converted from DVD to AVI (Lagarith YV12) using DVD2AVI.

Here you are some screenshots: top left image is raw capture from CLD-600; top right is the same capture, enhanced; bottom left is the average/median result of the three captures, enhanced, color corrected, resized to match GKAR; bottom right is GKAR, unprocessed. As I wrote before, the GKAR picture is always pinkish, while the laserdisc is pinkish only sometimes - at least in this side... obviously, GKAR has a better definition.

Starship is white on SET, pink on GKAR; planet is brown on SET, pink on GKAR; engine booster is right in SET.

Wall is white on SET, pinkish on GKAR; C3 is golden on SET, golden-ish on GKAR

Stormtrooper is white on SET, pink on GKAR

Darth Vader is black on SET, dark violet on GKAR

Leia's face color is more natural on SET, and whites are "whiter"

Sand is yellow in SET, orange-pink in GKAR

R2 is white in SET, pinkish in GKAR; the rocks color in SET is more similar to OT

I don't know if, in GKAR desert, they are sand waves, or MPEG waves... really, I'm not joking!

Maybe Luke is a bit pale on SET, but better pale than sunburnt as in GKAR!

The ground is brown in SET, violet in GKAR

Note RD's dome; still silver on SET, but what color is on GKAR?

C3 is golden on SET, pinkish on GKAR; in GKAR Luke's clothes are pink instead of white

Still working to improve the quality of the laserdisc captures; I'm almost sure that using three different editions (french, english, german) the noise will be lower; as the picture of 4300 and 600 is noisier and less detailed, I think I'll use only the 925 for all the captures - but I could also test the DVL-909, and maybe give a try on the Philips LDP-600WS...

OR, I could try the "extreme capture experiment (version 1)":

capture three times each edition; TOOT them to avoid drop-outs
repeat for three laserdisc editions
results are three TOOTed captures, one for each edition
average the three TOOTed captures
repeat for at least three laserdisc players
results are three averaged captures, one for each laserdisc player
median the three averaged captures

|capture 1| |->english-|capture 2|->tooted-| | |capture 3| | | | | |capture 1| | cld-600---|->french--|capture 2|->tooted-|->average-| | |capture 3| | | | | | | |capture 1| | | |->german--|capture 2|->tooted-| | |capture 3| | | |capture 1| | |->english-|capture 2|->tooted-| | | |capture 3| | | | | | | |capture 1| | | cld-d925--|->french--|capture 2|->tooted-|->average-|->median | |capture 3| | | | | | | |capture 1| | | |->german--|capture 2|->tooted-| | |capture 3| | | |capture 1| | |->english-|capture 2|->tooted-| | | |capture 3| | | | | | | |capture 1| | | ld-v4300d-|->french--|capture 2|->tooted-|->average-| | |capture 3| | | | | |capture 1| | |->german--|capture 2|->tooted-| |capture 3|

OR, I could try the "extreme capture experiment (version 2)":

capture three times each edition; TOOT them to avoid drop-outs
repeat for at least three laserdisc players
results are three TOOTed captures, one for each laserdisc player
average the three TOOTed captures
repeat for three laserdisc editions
results are three averaged captures, one for each edition
median the three averaged captures

|capture 1| |->cld600--|capture 2|->tooted-| | |capture 3| | | | | |capture 1| | english---|->cldd925-|capture 2|->tooted-|->average-| | |capture 3| | | | | | | |capture 1| | | |->ldv4300d|capture 2|->tooted-| | |capture 3| | | |capture 1| | |->cld600--|capture 2|->tooted-| | | |capture 3| | | | | | | |capture 1| | | french----|->cldd925-|capture 2|->tooted-|->average-|->median | |capture 3| | | | | | | |capture 1| | | |->ldv4300d|capture 2|->tooted-| | |capture 3| | | |capture 1| | |->cld600--|capture 2|->tooted-| | | |capture 3| | | | | | | |capture 1| | | german----|->cldd925-|capture 2|->tooted-|->average-| | |capture 3| | | | | |capture 1| | |->ldv4300d|capture 2|->tooted-| |capture 3|

totalling 27 captures (each 3 sides) = at least 54 hours of captures, and about 1.5TB of hard drive!

Any idea or experience to share?