logo Sign In

g-force

User Group
Members
Join date
24-Sep-2004
Last activity
16-Jul-2020
Posts
662

Post History

Post
#315075
Topic
GOUT image stabilization - Released
Time
Arnie.d said:


I also assumed the PAL version was cropped a bit more at the top and bottom. Apparently I'm also wrong about that.


Interesting. It looks to me like there are more lines at the bottom of the NTSC and more at the top of the PAL. And the PAL ROTJ does look a bit better in terms of ringing and detail. So, I'm gonna need a PAL GOUT set for Empire and ROTJ? ;)

-G
Post
#315028
Topic
GOUT image stabilization - Released
Time
Oldschooljedi said:

I'm doing encodings with this script. Source is PAL, here is the result of the encoding:
http://www.megaupload.com/de/?d=69Y24DX2

I'm really happy with the result of this, have not encoded more of the GOUT yet.

I would post some more encodings later, but I'm moving to another town in a few days and I won't have internetaccess for a while.


thanks for posting that Oldschooljedi! It comfirmed for me that the croping is a little off. It seems that NNedi causes some artifacts on the scanline right below the top border. I'm wondering if you crop with

Crop(8,4,-4,-2,align=true).AddBorders(6,68,6,66)

instead, if that would be better.

If you have the chance to try this and report back I'd appreciate it. Good luck with the move!

-G
Post
#314953
Topic
GOUT image stabilization - Released
Time
Moth3r said:

A 10-pass encode is totally over the top. The law of diminishing returns kicks in after 2 or 3 passes, and most people agree there is no benefit in using more than 4 or 5.


I've heard this from others also, but have no first hand experience. If Arnie.D has the time though, 10 passes at least doesn't hurt anything.

Not sure if anyone noticed, but I updated the script because I noticed some PAL resizing errors that I think I have sorted out now. Anyone in PAL land care to check them and report back?

-G
Post
#314809
Topic
GOUT image stabilization - Released
Time
seventiesfilmnut said:

QUick question G-Force: when encoding in VirtualDubMod which codec do you select?

Also do you leave the re-muxing of the audio till the last stage?

Many thanks!

- John


John, I honestly haven't gotten that far yet. Currently, I'm encoding the avs script to Mpeg with HCenc, and just limited portions of it so far. I plan on adding the audio back in when I re-encode the whole thing to DVD.

It would be cool to have a huffy file of this though so that non-linear editing could be done later. I don't know though, something about manually painting frames doesn't sit well with me. Maybe if someone wanted to use this for replacing scenes from the SE, but then I think I'd suggest applying the filters differently depending on the scene. The global motion stabilization for instance can be so much better if you know there are no large objects moving in scene (and even better yet if there is no intentional motion at all). And I'm also toying around with making the stars less shimmery, but am having trouble doing it safely on the whole movie.

So that's the long winded answer to your question about encoding, which is really just me thinking out loud about the need for a lossless file.

-G
Post
#314785
Topic
GOUT image stabilization - Released
Time
Moth3r said:

...his AviSynth scripting is more advanced than anything I could do.


[blush]Come on[/blush] there was a time just a couple months ago where everything I knew was from you. I would dissect your scripts and run parts of them just to see what they were doing.

Moth3r said:

However, I've given up on the GOUT discs for the time being.

I can't blame you there. I figure this is the best we have so far, so I want to make it as good as I can and then I'll have the skills to do the next version that needs fixing. And in the mean time, if I ever stop tinkering and actually encode the whole thing, I'll have something watchable.

Moth3r said:


I'll probably be taking a look at laserdisc sources in the future.

I can't wait!

BTW, Small updates to the script on page one. No change to the picture, but should speed things up a bit.

-G
Post
#314589
Topic
GOUT image stabilization - Released
Time
Oldschooljedi said:

It seems, that there was something wrong with the EEDI2-plugin.
When I did encodings with "NNEDI(dh=true,field=1).TurnRight().EEDI2().TurnLeft()" the lasers of the imperial starship over Tattooine sometimes turned from green to white and looked a littlebit like bubbles.

Now I tried your script from the 20.03. with "NNEDI(dh=true,field=1).TurnRight().NNEDI(dh=true,field=1).TurnLeft()" and everything works fine.


Oldschooljedi, thanks for spotting this! I found the problem you are referring to, and it turns out that the error isn't in the EEDI2, it's in the motion compensation stage. I added "thSAD" to the MVCompensate lines, which fixes this problem. (see updated script on first page)
I've also moved the resized to before the TurnLeft, which speeds up processing a bit.

-G

Post
#314558
Topic
GOUT image stabilization - Released
Time
Oldschooljedi said:

I have the same error at my encodings.

G-force, can you remember what the error was?

I have this error when I use HC encoder, not with other encoders. But HC encoder runs faster with this script than other (commercial) encoders, so I prefer this encoder.

By the way: What version of avisynth do you have installed and what encoder are you using?



I was having that problem with QUenc, but then I switched to HCenc and the problem went away. Strange, seems to be the exact opposite of your problem.

I am using avisynth2.5.

As for the double NNEDI, The second one doesn't do anything. I'll check the lasers you mentioned and update the script as needed.

-G
Post
#314278
Topic
GOUT image stabilization - Released
Time
seventiesfilmnut said:

Trouble is the ringing is alot more noticeable in the 2nd and 3rd examples. Personally I think I'd rather stick with the original for this scene. I guess it's all a huge balancing act, and there isn't one script that can do the entire film - I'm leaning towards splitting the film into scenes and using different filters depending on what each clip dictates.


I like the idea of splitting up each scene. Would take a lot of work though. Sticking with the original for this scene is not an option in my opinion, as there is was too much crazy local motion (things that shouldn't be moving are)

Anyway, it turns out that the script that I used to generate #2 had too much sharpness, and as a result the playback appears to "nervous". It's starting to re-introduce the local motion that I'm trying to remove.

Also, I've got a new star preservation routine that I'll post into the script on page one.

-G
Post
#314237
Topic
GOUT image stabilization - Released
Time
Arnie.d said:

Are you sure the vertical anti-aliasing is necessary? I think the vertical aliasing is barely noticable and if you remove it you blur the image more.


The vertical aliasing is VERY visible in VERY rare places. So I'm torn. A little bit more sharpness for the whole movie, and faster running script, or fixing some very rare but extremely ugly problems? The issue I'm talking about can be seen here:

[IMG]http://img87.imageshack.us/img87/4302/owenorigxh0.th.jpg[/IMG]
[IMG]http://img72.imageshack.us/img72/2860/nnedily5.th.jpg[/IMG]
[IMG]http://img87.imageshack.us/img87/9493/nnedieedi2ce3.th.jpg[/IMG]

Make sure to click on the image after it opens to see it full size. The 2 anti-aliased ones were made without changing any sharpening settings. Notice the zaggies on the droid on the right that are fixed by the rotated EEDI2. What does everyone prefer?

-G
Post
#314187
Topic
GOUT image stabilization - Released
Time
seventiesfilmnut said:


I also downloaded a different mt_masktools.dll (dated 27th Jan '06) from http://manao4.free.fr/


Oops, I forgot that that .dll is a bit tricky as well. I think I'm using Masktools v.2. Not sure of the date, but most likely at lease 2.0 alpha 27 which is the most recent listed on this page: http://manao4.free.fr/mt_masktools.html
but I noticed he also has some higher versions on the page that seventiesfilmnut listed.

I kinda had to go through this same process of aquiring the necessary plugins when Arnie.d and Moth3r started the anti-aliasing thread. I wouldn't want to deny you guys the same fun! ;)

-G
Post
#314105
Topic
GOUT image stabilization - Released
Time
Arnie.d said:

I have found some more scenes in the film were eedi2 causes serious artifacts


That's it. The final nail in the coffin for EEDI2. I also found a blur bubble on a stormtrooper helmet. I sure liked how it looked over NNEDI on the R2 scene, but I can't justify screwing up everything else as a result. Plus NNEDI blurs way less.

Updating script now to switch to NNEDI.

seventiesfilmnut said:

Hi G-Force!

Been lurking on this thread for a while and just wanted to say it's great the work you've put into this - looks fantastic.

Edit: Just found out about DGIndex which I'm using now to create the D2V files - playing catch-up here :) Using the 64-bit floating option under video - is this correct?


Thanks for the kind words! I'm glad you figured out the DGindex thing. Not sure about the 64bit thing right now. Will check at home. Sorry I didn't respond sooner. So far, I've only done the full script on parts of the film, but only because it's getting really slow! I'm sure there will be some tweaks after the first full encode.

-G
Post
#313974
Topic
GOUT image stabilization - Released
Time
I checked your conversion to PAL resolution, and it keeps it close enough to the correct dimensions. It may not run as quickly though, since everying before the resize is MOD2. You might try:

orig=Crop(0,120,0,-120,align=true)
...
Spline36Resize(720,448)
...
AddBorders(6,64,6,64)

I'll check all the pages you posted for the plugins and verify against mine tonight when I get home. In the mean time, make sure you have Removedirt from here: http://home.arcor.de/kassandro/RemoveDirt/RemoveDirt.htm
and Removegrain from here:
http://home.pages.at/kassandro/RemoveGrain/

-G

Post
#313944
Topic
GOUT image stabilization - Released
Time
Here's a pic of the same scene through my most recent script (moved levels to after temporal sharpener)

http://www.wackychimp.com/uploads/HanGreedo.jpg

I agree with you Arnie.d, I think the grain adds the perception of detail also, but it was a somewhat unintended casualty of the local stabilization that I did. Oh well. ;)

Anyway, I have updated the script on page 1, and it now reflects what I used to generate the pic in this post. I have not yet annotated the script, and I have some more things that I would like to do, but it's there for your viewing pleasure.

-G
Post
#313911
Topic
GOUT image stabilization - Released
Time
Arnie.d said:

I have tried out so many filters and as a result what was in my memory didn't match with the actual original. I included my own for comparison. Yours is much cleaner no doubt. I think mine is a tiny bit sharper (but also darker and more grainy).


Ugh. I was hoping you wouldn't point out that scene. For the last clip I posted I was doing the motion compensation/sharpening after my levels command and I noticed that some softening of the picture was happening in the darker scenes. Moving the levels command to the end helps this a lot, but I still think I can do a bit better. I have an idea that will mitigate the blurring effects of the eedi2 command, then when the temporal sharpener comes in, there will be more detail to work with. Yours is really nice and sharp! Is that even after the eedi2? What are you using to sharpen? Do you think the lighter scenes in mine are too soft also? (like Leia standing in front of the stormtroopers?)

-G
Post
#313890
Topic
GOUT image stabilization - Released
Time
Arnie.d said:


I was wondering, although the image looks pretty clean don't you think it's a bit to soft?

Also, do you use fft3d for noise filtering or sharpening (assuming you still use it).


Yes, I'm still using FFT3D for both filtering and sharpening at one stage (although less than before). Would you mind posting some comparison pics of mine compared to the original showing me what you mean by too soft? I personally don't see it, but I may not be looking in the right place.

thanks,

-G