Sign In

GOUT image stabilization — Page 3

Author
Time
Originally posted by: pittrek
So which script is the actual one ?


the one on the first page.

-G
Author
Time
New test encode.

stabilize.m2v

Please take a look and tell me what's wrong with my Tie lasers? This problem doesn't show up in the Avisynth output, it is something that is happening during the encode. Any ideas?

-G
Author
Time
Moth3r,

I may have figured out why there is a discrepency between what I think looks like a good hue and what you used. Also, this would tell me that when I encode it, it's going to be darker (more reason for my added gamma correction) Discussion can be found here:

http://forum.doom9.org/showthread.php?s=&threadid=82217

What I'm wondering now is if your screencaps on the fanedit site are from the Avisynth script or from the encoded DVD. I'm assuming the encoded DVD, but I want to make sure.

I'll play around with the Colormatrix() plugin and try some test encodes.

-G
Author
Time
Script updated with even better image stabilization (see 1st page). Clip coming soon.

-G
Author
Time
Well it looks nice, but it's still pretty shaky.
Could you (if you have time) do again a comparision video ?
Author
Time
Originally posted by: pittrek
Well it looks nice, but it's still pretty shaky.
Could you (if you have time) do again a comparision video ?


It is still a little shakey in localized areas. I'm still working on that.

Here's a comparison clip with the latest script. (bottom is original)



stabilize_compare.m2v

-G
Author
Time
I've had some MAJOR breakthroughs in the last couple days with respect to local motion stabilization and dirt removal (results were so positive that I even wrote a script to share with the helpful folks at the Doom9 forum). This thing is rock solid. Stay tuned for some clips. I've also found out how to repair a TON of detail by collecting pixels from the same object in different frames. The results are a little preliminary (like as of before work this morning) but the results so far are breathtaking (I literally couldn't believe my eyes!).

This is really going to give the X0 project a run for it's money!

Like I said, stay tuned!

-G
Author
Time
Sounds good, and it's great to see someone become enthusiastic about AviSynth!

However, you're not going to get results to rival the X0 project (IMHO, of course) unless you figure out a way to remove the "motion smearing" caused by the THX process DVNR. The X0 project got around this by using pre-THX laserdisc releases.

Didée mentions a "temporal sharpener" in this thread, but I don't think a working filter was ever produced.

Guidelines for post content and general behaviour: read announcement here

Max. allowable image sizes in signatures: reminder here

Author
Time
Originally posted by: Moth3r

However, you're not going to get results to rival the X0 project (IMHO, of course) unless you figure out a way to remove the "motion smearing" caused by the THX process DVNR. The X0 project got around this by using pre-THX laserdisc releases.

Didée mentions a "temporal sharpener" in this thread, but I don't think a working filter was ever produced.


Agreed! There is no way to remove all of the motion smearing from the GOUT. I wish I had a raw X0 capture of the pre-THX version to work with. I would never claim that the GOUT could ever be better than the X0 in that respect (but maybe in the parts that aren't smeared!?!?!).

Didee has let enough ideas out about his "temporal sharpener" that it looks like I can incorporate some version of his idea it into my local-motion compensation algorithm. More to come....

-G


Author
Time
It might rival the X0 project if they were not using the pre THX laserdiscs as well. Still I hope they are aware of the stuff you've done because if it restores detail the script might be of some use to them. Maybe you should see if you could compare some of your results.

Is there some script out there that could combine the vertical line resolution of the Pal laserdiscs with the horisontal detail of the Gout? Like know to use the NTSC scanlines but whereever there's an additional Pal line in between drop it in.

Take back the trilogy. Execute Order '77

http://www.youtube.com/user/Knightmessenger

Author
Time
Originally posted by: Knightmessenger
It might rival the X0 project if they were not using the pre THX laserdiscs as well. Still I hope they are aware of the stuff you've done because if it restores detail the script might be of some use to them. Maybe you should see if you could compare some of your results.

Is there some script out there that could combine the vertical line resolution of the Pal laserdiscs with the horisontal detail of the Gout? Like know to use the NTSC scanlines but whereever there's an additional Pal line in between drop it in.


They may be able to do some of my stuff post process. I'm sure they are far enough along that they wouldn't want to do it to the raw capture. Here's to hoping that the X0 is even better than what I finally come up with.

It's a good idea about combining the PAL LD and the GOUT sources in some way. I have some ideas for this, but don't have the PAL capture to work with. [cough] new rip [/cough]

-G

Author
Time
I won't be able to post the script that I used to make this for a couple days, since I just threw it together and want to clean up the script a bit first. Here's a new clip though. Let me know what you think.

http://rapidshare.com/files/98121037/stabilize_hc2.m2v.html

-G

Author
Time
Wow!

I did author your testclip to DVD and watched it on my widescreen-TV. This looks really great. I can't wait to see the script you used for this!
Author
Time
Were you not as vigorous with the stabilization script for that encode?

“I love Darth Editous and I’m not ashamed to admit it.” ~ADigitalMan

Author
Time
Originally posted by: g-force
I won't be able to post the script that I used to make this for a couple days, since I just threw it together and want to clean up the script a bit first. Here's a new clip though. Let me know what you think.


I saw your clip, and the various sequences look pretty tight--except for the title cards. Thus, I am wondering if it would make more sense to re-do these from scratch. You are trying to stabilize elements that can be reproduced fairly easily on a computer--and with ZERO shaking.

The "A long time ago..." card is a piece of cake; you need only ONE source frame to faithfully reproduce it. The title sequence is a bit trickier, but can be reproduced quite accurately in AfterEffects provided that the font is really NewsGoth.

If you are interested in pursuing this, let me know.
Track the Star Wars Prologue fanedit at
Digital Fanedits
Author
Time
Originally posted by: Oldschooljedi
Wow!

I did author your testclip to DVD and watched it on my widescreen-TV. This looks really great. I can't wait to see the script you used for this!

I'm glad you like it. I just realized that really fast motion stuff is too blurred though. I will keep working on it. In the mean time, I'm leaving the script as is, since it's the best so far without adversly affecting the quality in any way.

Originally posted by: boba feta
Were you not as vigorous with the stabilization script for that encode?


The last clip is actually the most vigorous that I've been, but since I'm applying the same script to the whole thing, some stabilization may be better in spots and not as good in others. I could do even better if I had different settings on each scene (less artifacting in some, better satabilization in others). I would like to get it the best I can for the whole movie first so that there are only a few select scenes that I need to change the settings.

Originally posted by: Anti-Matter

The "A long time ago..." card is a piece of cake; you need only ONE source frame to faithfully reproduce it. The title sequence is a bit trickier, but can be reproduced quite accurately in AfterEffects provided that the font is really NewsGoth.

If you are interested in pursuing this, let me know.


The "A long time ago..." is definitely one of the things that can be better easily using Avisynth, the scroll will definitely be trickier. I really wouldn't mind a good reproduction of the post 1979 scroll, since I don't think of this as the original theatrical release anyway. So yeah, I'm interested!

-G





Author
Time
I really wouldn't mind a good reproduction of the post 1979 scroll, since I don't think of this as the original theatrical release anyway.


So you want the lime-yellow version without an episode number? If so, the logo and velocity can be sourced from the bonus DVD.

If someone can provide the following elements, I will try to re-build the original crawl for you.
1. The first complete frame of the SW tile sequence (from the bonus DVD, which I do not have);
2. Your opening starfield (or the cleanest frame after the text fades out)
3. Your pan-to frame (ideally, the frame immediately before the fly-in).

Thus, I need only three PNG files to get started.
Track the Star Wars Prologue fanedit at
Digital Fanedits
Author
Time
I was actually thinking of the version with the Episode number (since that is the era that the GOUT was originally mastered)

With that in mind, would you just use the crawl from the Special edition, and replace the starfields? The end credits might benefit from something like this as well.

-G
Author
Time
It doesn't really matter; but you have to first make some decisions, because you cannot arbitrarily choose starfields (unless you want the title sequence to look amateurish).

You *must* use the starfields from your existing video elements--especially the pan-to frame. The typography is completely independent, except that you presumably want it to match a specific version of the movie. But which version?

I was assuming that any effort to restore the original version would also try to restore the 1977 crawl. Your clip uses the 1977 logo and color.
Track the Star Wars Prologue fanedit at
Digital Fanedits
Author
Time
Anti-Matter said:

but you have to first make some decisions


But I've been decisive all along. ;) Post 1979 scroll (the one with the episode number). The Greedo subtitles are not original, so the opening scroll shouldn't be either.

-G
Author
Time
 (Edited)
New test encode. This one has no loss of detail on fast moving objects:

http://rapidshare.com/files/98695507/stabilize_hc3.m2v.html

and for comparison, the original resized to anamorphic:

http://rapidshare.com/files/98705782/original_resized.m2v.html

Please share your thoughts.

-G
Author
Time
g-force,

could you post your script you've been using for your previous 'stabilize_hc2.m2v' encoding?

The image looked very good on my widescreen-tv and I would like to do some PAL-testencodings.

The colours look more theatrical to me on this encoding.

Author
Time
New test encode. This one has better local stabilization, less artifacting (actually none that I can see), and temporal sharpening!

http://rapidshare.com/files/99217835/stabilize_hc4.m2v.html

This may be as good as I can get this, so the script is going up as soon as I can annotate it.

Let me know what you think!

-G
Author
Time
I think that's pretty damn good actually, but I wouldn't mind seeing a few selected scenes considered to be some of the worst i.e: the stormtroopers finding the O Ring by the escape pod, R2 in the canyon and the landspeeder entering Mos Eisley. I think they all suffered terribly from the DVNR process.

“I love Darth Editous and I’m not ashamed to admit it.” ~ADigitalMan