logo Sign In

.: The Zion DVD Project :. (Released) — Page 10

Author
Time
Post link
I think the halo-problem can't be solved through chroma-shifting, since it's a BRIGHTNESS-problem... I guess the exorcist-filter works by multiplying an inverted version of the image on top of the original (but only in the overlapping parts, caused by the given offset). This results in bright areas getting darker, and dark areas getting brighter, which in most cases is fine. Unfortunately, some dark areas (like those which were black before) don't need to get brighten up, and so this causes some errors.

I have found a little workaround for this through AviSynth, although by this workaround only the bright-halos get removed. But since these are the most visible ones, I think it might be an option. I'll post the script and some screenshots later.
Author
Time
Post link
could I just say two things Zion from two days of tweaking and fiddling with my own ANH?

although I lost my original capture, I 've been playing around with the MSU sharpen and denoise filters on my original empire capture and what I have found is that

a) MSU filters are bloody slow. my works 1.7athlon is only yielding 1 or 2 fps when either of these filters is present. with any other sharpen filter I am getting 14fps so that's a massive hit. based on that it will take my system 49 hours to give me an interim file which is ludicrous.

b) the MSU sharpen filter, while pretty good, introduces a certain amount of mosquito noise to the source.

in comparison if I use unsharpmask.vdf under virtualdub I get nearly as good a sharpness as MSU but with less noise. there is still some, but not as much as MSU.

I find it interesting that you IVTC in virtualdub rather than in script. I found I had field issues whenever I used virtualdubs own. on the other hand, avisynths decomb filter by donald graft has always proven to be very good.

you haven't mentioned it but did you ever manage to eliminate the jaggies? i have them present on my capture but have yet to find a filter that can help on that. some kind of dot crawl thing perhaps? any suggestions?

finally, you seem to be capturing awfully dark? I know this can be adjusted in post but surely you are losing some small bits of detail in the original dim capture? it also explains how it came out a bit dark on your TV. I had this same problem and ended up compensating a bit too far the other way. ie, the blacks were a bit washed out. its tough I know and usually involves a few passes before you hit the right mix to suit.
When a woman says yes, she means no - when she says maybe, she means no.

http://www.auky37.dsl.pipex.com/falconlogo_web.jpg
Author
Time
Post link
on another note regarding your menu's
very good. impressive, most impressive... sorry, anyway, what I would say in response to the comments some people made saying there was too long between the bits of dialogue, why not just put the Imperial march on in the background. that would fill the quite moments, plus it's appropriate piece of musice for the menu theme you are using.

that way there are no quiet moments.
When a woman says yes, she means no - when she says maybe, she means no.

http://www.auky37.dsl.pipex.com/falconlogo_web.jpg
Author
Time
Post link
tellan,

Yes, the MSU filters are ungodly slow. If not for the smart sharpen's efficiency, I wouldn't even have given it a second look. However, I had great results using it along with the AVIsynth convolution3d filter.

As far as IVTCing with virtualdub, you have to start it on the first progressive frame (P P P I I P...) in order to make it work. Otherwise it doesn't work at all. I haven't seen any interlaced frames sneak by with this method.

No, I have not eliminated the jaggies. If you come across a good filter for this, please let me know. I haven't even begun to look for one yet.

As for the darkness of my capture, I do that intentionally in order to not lose the bright details. If you look at that first screencap in the *Official* screencaps thread, you will notice that some of the other captures are way too bright, and the details in the whites are all but gone. In capturing a little darker, I am able to retain a little of that detail. Of course this is at the cost of a little detail loss on the dark end, but I feel this is a compromise I'm willing to make - at least for this film. I'm thinking ESB and ROTJ won't need this so much.

My Projects:
[Holiday Special Hybrid DVD v2]
[X0 Project]
[Backstroke of the West DVD]
[ROTS Theatrical DVD]

Author
Time
Post link
I see what you're saying about the white values. mine suffers from the same issue as the others. the white corridors of the tantive have blown out a bit right after the interogation scene when vader and the black shirt are walking down the corridor. also I lose the top of the stormtroopers helmet in the ceiling light.

re ESB and ROTJ you won't have this problem. it's only ANH where the first 20 mins or so are soft and the colour values are a bit off. once they get to mos eisley, things improve immeasurably which rather illustrates the parlous state of the print the LD's were struck from at the time prior to be digitally remastered in the mid 90's

as regards jaggies, no I don't have an answer I'm afraid. I've found this to be a combination of factors

1. the comb filter on your capture card. mine gave me dot crawl which I cannot eliminate and there are a certain amount of jaggies.
2. which in turn is made worse slightly when you IVTC back to 23.97fps as the frame decimation and field blending creates some more
3. which in turn is exaggerated slightly more when you stretch it for anamorphic

there isn't really an answer to that one I think. it basically comes down originally to how good your capture card is. but they all suffer from it to some degree.


another thought regarding your dark capture. if you're using that toot filter that allows you to directly integrate a number of captures into a single interim, couldn't you capture one light and one dark and combine the two? that way you get the highlight and shadow detail but also keep high white details that would otherwise be lost. course on this point I could be talking out my arse.
When a woman says yes, she means no - when she says maybe, she means no.

http://www.auky37.dsl.pipex.com/falconlogo_web.jpg
Author
Time
Post link
All right, time for some screenies regarding my "new" ghostbusting-method. The old method (using the Exorcist-filter in VirtualDub) caused some new halos, so I guess it wouldn't work that way. Here's a comparison between the old and the new one:

http://www.scummbar.com/mi2/temp/oldnew3.gif

On first sight they look similar, but let's take a closer look, by brightening up the image:

http://www.scummbar.com/mi2/temp/oldnew1.gif

As you can see, the old method caused quite an ugly halo to appear left of Leia's face, which of course resulted from brightening up that "halo-area" (of course there was no halo visible in the original source, instead the color was SUPPOSED to be dark there... but how could the filter know that?). So I thought it was necessary, to not brighten up dark halos, and ONLY darkening the bright halos, which worked very well through AviSynth, by layering an offset version of the image on top of the original, using the "Darken" transfer-mode.

All right, we've got rid of the bright halos, and there's no "new" halo on the left side. But, as you might notice, there's now a dark-halo on the right side on Leia's hair. OK, this halo IS supposed to be brighten up, but we can't do that, because... well, have a look at the next image:

http://www.scummbar.com/mi2/temp/oldnew2.gif

You see that black part on the Stormtrooper's arm? If we would use a filter to brighten up the halo around Leia's hair, the black part on the stormtrooper would get a bright halo... something we don't want (it's looking good in the original). But since it's only in the dark parts of the image, i guess we could leave it this way.

Here's the AviSynth-script:
    v1=AviSource("e:\zion.avi").ConvertToYUY2.GuavaComb(mode="NTSC").SeparateFields.SelectEvery(10,0,1,2,3,4,5,6,9).Weave.AssumeFPS(23.976)

    v2=v1.crop(4,0,0,0).AddBorders(0,0,4,0, color=$FFFFFF).blur(.5)

    v3 = overlay(v1,v2, opacity=0.2, mode="Darken") # remove only bright halos

    return v3

I guess now it's necessary to IVTC in AviSynth, since the image gets blurred in line 2. This requires the video to already be "IVTC"ed, because blurring a video with fields is basically suicide.
Author
Time
Post link
Originally posted by: tellan
on another note regarding your menu's
very good. impressive, most impressive... sorry, anyway, what I would say in response to the comments some people made saying there was too long between the bits of dialogue, why not just put the Imperial march on in the background. that would fill the quite moments, plus it's appropriate piece of musice for the menu theme you are using.

that way there are no quiet moments.


I reckon U should leave the audio as it is. It sounds spooky. I love it.

E Chu Ta!

Author
Time
Post link
Originally posted by: zion
So you're not using any of the Virtualdub filters anymore? Is that script the only thing you're doing to the video?


I am frameserving that script into VirtualDub where I crop and resize the image. And other than that sharpening-approach I showed in my "filter overview" a few posts ago (through upscaling, Xsharpen and downsizing again) I am not doing anything else in VirtualDub. Although, maybe that sharpening isn't necessary. The downside of my new ghostbuster is that there's a little more noise left than with my old method. But again, I don't mind the noise... it makes the image look more natural.

Oh, and my IVTC-method (using the SelectEvery-command) only works with your sample (or any clip that starts on the same frame of the pulldown-cycle). I guess the Telecide-filter for AviSynth would yield the same result AND it would recognize the pulldown-cycle correctly.

Author
Time
Post link

Ash, it does look a little over saturated. Maybe I need to tone it down a little.


Maybe not just a bit oversaturated, but also it seems like there is a lot more detail lost between those two versions. Like the lines on 3P0's leg or the reflections in the blue parts on R2's dome. I don't know how much smoothing or noise reduction is necessary to eliminate artifacts introduced during capture, but personally, I'd rather keep as much of the original laserdisc picture detail as possible, even if that means some more noise.

Author
Time
Post link
I think the detail-differences are just caused by the two images not being the same frames in the movie. As you can see, 3PO isn't as close to the camera as in the upper image, and so I guess reflection on R2's blue parts occures in this later (or earlier) frame.

But it's true, the images are still a little over-saturated. This is especially visible on these screenshots:

http://www.aptirrelevance.com/videoprojects/images/ANH.jer.dvdi.04.jpg

http://www.aptirrelevance.com/videoprojects/images/ANH.jer.dvdi.23.jpg

They look kinda sunburned... but with a little less saturation your color-correction looks just great.
Author
Time
Post link
Oh no... I have to disappoint you guys again... my "new" ghostbuster-method wouldn't work either. I have just tried it on some other pics, and this is the problem:

http://www.scummbar.com/mi2/temp/anhx.jpg

The new method only takes care of ghosts being cast FROM darker areas ONTO brighter areas, resulting in even brighter halos.

It does NOT take care of darker halos being cast FROM brighter areas ONTO darker areas... as you can see the bright twin-suns are casting dark ghosts onto the sky... unacceptable. Maybe my old method is better? No... well... the bright halos in the dark areas (check the stormtrooper-arm above) aren't acceptable either... arrgh!!!
Author
Time
Post link
Dammit. Well hopefully we can come up with something else. I'm going to at least try your new method and post some color-corrected (less saturated) screencaps just to see what it looks like.

One thing I noticed yesterday while playing around with my color correcter settings was that there was a definite threshold I could not cross with brightness and contrast because the video noise would become a lot more obvious.

Right now I'm thinking of upgrading my hardware. I have a feeling that the ghosting and the noise have a lot to do with the signal flow from the LD player to the capture card. One thing that annoys the hell out of me is the cheap multi-connector that I have to plug my video cable into. It's not a jack right on the capture card, but a crappy cable with a din connector on one end that splits off into audio, svid, and composite. This pretty much renders my Monster video cable useless, as the video transfer is only as good as the weakest link. I'm also noticing that the video out on my LD player is loose, and loses the signal completely if the jack is moved around. This can't be good either.

So I am now on the lookout for a new capture card and LD player. Of course I would love to use a CLD-97 or CLD-99 or even a CLD-D704 for this project. Funds are pretty tight right now though, so here's hoping I can find a good deal or someone to let me borrow one.

The capture card I have my eye on is the MSI TV@nywhere Master. It's a simple, cheap, high quality card with the same chip as my current card, only it has input jacks directly on the card.

My Projects:
[Holiday Special Hybrid DVD v2]
[X0 Project]
[Backstroke of the West DVD]
[ROTS Theatrical DVD]

Author
Time
Post link
I am trying to find a solution for removing the ghosts in a NICE way... although trying to capture a ghost-less image would of course be even better. Have you ever checked out the laserdisc on a quality display... are there any ghosting artifacts already on the LD?
Author
Time
Post link
Something new on the IVTC-issue:

Get the DeComb-package, and then you can just use the Telecide(1) command in AviSynth. So you no longer need to separate and weave the fields manually. In the script this would be enough:
    AviSource("e:\zion.avi").ConvertToYUY2.GuavaComb(mode="NTSC").Telecide(1)

    AssumeFPS(23.976)

    return last
I don't know if the AssumeFPS command is really necessary, but without it the resulting clip only has a framerate of 23.970 ... which, I think, wouldn't work when creating a DVD out of it. But you should try it, and check, if the audio is in sync.
Author
Time
Post link
http://www.aptirrelevance.com/videoprojects/images/ANH.jer.dvdi.23.jpg
http://home.netcom.com/~eandtc/sitebuildercontent/sitebuilderpictures/han.jpg

I compared this image with the Dr. Gonzo transfer, and it looks like you are filtering out too much horizontal information. On the instrument panel on the right side of the screen, alot of the white lines have been simply removed!

aka nostromo777 on myspleen

Author
Time
Post link
Originally posted by: zion
Yeah, I just saw your email about that too. You've hereby convinced me that I'm capturing too dark. Time for some more tweaking....


Yeah, and do something about Harrison's plaque.
The Jedi are all but extinct.......
Author
Time
Post link
zion,

In the two R2 comparison shots the old one still looks sharper. Personally I'd rather not sacrifice any sharpness to remove halos that are really hard to detect.

Thant's my two cents
The Dean
Author
Time
Post link
http://home.netcom.com/~eandtc/sitebuildercontent/sitebuilderpictures/han.jpg

zion, I'm finding out that it's better to capture with as neutral values you can get, and then playing with color/contrast/etc. in your video editor. This pic is from my raw capture. Nothing really sticks out, but every color and detail is visible.

<span class=“Italics”>MeBeJedi: Sadly, I believe the prequels are beyond repair.
<span class=“Bold”>JediRandy: They’re certainly beyond any repair you’re capable of making.</span></span>

<span class=“Italics”>MeBeJedi: You aren’t one of us.
<span class=“Bold”>Go-Mer-Tonic: I can’t say I find that very disappointing.</span></span>

<span class=“Italics”>JediRandy: I won’t suck as much as a fan edit.</span>

Author
Time
Post link
guys, I thought I'd share this as a possibility for you to consider

go to

http://pcvideo.tripod.com/

I was tooling around today trying to find a sharpener for virtualdub. MSU is far too slow to counteract the benefit, plus I think it introduces too much video noise into the capture. unsharp mask is not bad, not as good as MSU but too much edge enhancement, then I came across a filter on this site.

you get absolutely zip control of the filter selections, you get the default they think best, but my god it hasn't half improved mine.

I used

gentle sharpen,
saturation

and have seen a marked improvement. it almost acts like auto sharpen and auto levels in photoshop.

the sharpen in particular does an excellent job with next to no mosquito noise being created.

so the filters I use are this new filter to sharpen and saturate and donald grafts RGB filter from here http://www.neuron2.net/ to finely adjust the red bias and give a slightly better white.

I can post people a screenie or two if they're interested. my FTP isn't working so I can't put them up on here.
When a woman says yes, she means no - when she says maybe, she means no.

http://www.auky37.dsl.pipex.com/falconlogo_web.jpg
Author
Time
Post link
Echo3 stole my pic!

[EDIT]: I just realized that I lied. That isn't my raw capture, it's from my IVTC'ed file, with a dot-crawl filter applied. Other than that, nothing is changed.

<span class=“Italics”>MeBeJedi: Sadly, I believe the prequels are beyond repair.
<span class=“Bold”>JediRandy: They’re certainly beyond any repair you’re capable of making.</span></span>

<span class=“Italics”>MeBeJedi: You aren’t one of us.
<span class=“Bold”>Go-Mer-Tonic: I can’t say I find that very disappointing.</span></span>

<span class=“Italics”>JediRandy: I won’t suck as much as a fan edit.</span>

Author
Time
Post link
sorry, MeBeJedi, just wanted to make my point more obvious!

aka nostromo777 on myspleen