Not doing any cropping yet.  That is way down the road.

Regarding the frame borders, I tried to keep the camera in the same position for every scan.  On reel 1, two of the three scans have virtually identical border positions, with the third scan not matching quite as well.  As a result, I'm trying to use the two matching scans whenever possible, resorting only to the third scan when absolutely necessary.  If I can do that throughout, it will make it much easier later on.  I think that for each of the reels I have two good scans with identical camera placement (and different brightness, of course).

For assembly, I'm erring on the side of detail in dark and light scenes, trying to get less white/black blowout than I sometimes have in the past.

After assembly I assume I'll start in on color correction.  My understanding is that it is best to do that before resizing, right?  That will be slow, too, as I'll need to do it scene-by-scene as well.  My goal with color correction will be to stick as closely to what the print looks like (to me), rather than matching the GOUT or other sources.  As before, I'll err on the side of good skin tones where a decision has to be made.  If there is anything any of you think is of paramount importance to achieve in some particular scene, let me know now so that I can pick the best source from which to work on that scene... I tend to not pay attention to some of the details that people fret about here (like lightsaber colors), and so I wouldn't want to miss something key, unnecessarily.

One interesting thing I've noticed is that the entire ESB movie has surprisingly few green things in it.  There is some in Yoda, a little bit in the Hoth command center, and some in the foliage behind him, but that's about it. Even the cockpit and techie lights are almost all red and blue. It has made it difficult to get good starting white balance.  But that's nitpicking because the print and the capture looks great.

Quick question - why does R2 look black-and-white in some space scenes?  I thought it was an error on the print until I confirmed it in Adywan's reconstruction.

Puggo - Jar Jar's Yoda said:

Quick question - why does R2 look black-and-white in some space scenes?  I thought it was an error on the print until I confirmed it in Adywan's reconstruction.

I think the "blue" model conflicted with the bluescreen backgrounds.  Like the tv weatherman who wears a blue tie and suddenly has a map of Iowa projected onto it :-)   I'm open to being corrected.

none said:

When you say assembly, what does that mean exactly.  Are you looking through each of those scans for the 'best' quality (frame placement, consistant lighting, x, y, z), or are you averaging the scans as someone had recommended previously?

I actually tried some experiments with averaging the scans in Vegas.  It was pretty cool as it does manage to do a pretty good job of balancing dark and light areas of the frame.  But I couldn't get the frames to line up consistently enough.  If I took my closest two scans and superimposed them, I could make individual frames line up well, but then 3 or 4 frames later one would move a slight amount and it would lose sharpness.  Without a tool, this would take forever.  Is there a tool to help with this?

So, instead, I'm slicing it up scene-by-scene, taking each scene from whichever scan has the most detail based on which scan has the most appropriate aperture setting for that scene.  Later, I'll go through and individually color adjust them so they match as well as I can.

Hi Puggo,

I'd first like to say thanks for preserving these films so that they are available to all of us.

I noticed while reading this thread that you are looking for a better way to auto-align the frames from several of your reel captures.

I know of an avisynth script that should do a better job of auto-alignment than photoshop or vegas.  I've never used the vegas method but from what you are describing it's not that great. 

If you are interested just let me know what resolution your captures are and I can edit the script to fit the resolution and then post it here in this thread for you.

althor1138 said:

I noticed while reading this thread that you are looking for a better way to auto-align the frames from several of your reel captures.

I know of an avisynth script that should do a better job of auto-alignment than photoshop or vegas.  I've never used the vegas method but from what you are describing it's not that great. 

If you are interested just let me know what resolution your captures are and I can edit the script to fit the resolution and then post it here in this thread for you.

Thanks!  654 x 480

To clarify, I was only using Vegas to composite the frames. Aligning I was trying to do by hand.

Sounds fun to try.  I'm curious how good of a job it can do at aligning frames when the brightness levels are different.

**EDIT**

See post #139 for newer,better version

reel1cap2 = avisource("PATH TO FILE") #this is the video that will be altered.
reel1cap1 = avisource("PATH TO FILE") #this is the video that will be the source that the other is matched to.

colw = reel1cap2.Width()
colh = reel1cap2.Height()
bww = reel1cap1.Width()
bwh = reel1cap1.Height()
borw = (bww-colw)/2
borh = (bwh-colh)/2
reel1cap2= reel1cap2.AddBorders((654-reel1cap2.Width)/2, (480-reel1cap2.Height)/2, (654-reel1cap2.Width)/2, (480-reel1cap2.Height)/2, color=$000000)
bwp = reel1cap1.AddBorders((654-reel1cap1.Width)/2, (480-reel1cap1.Height)/2, (654-reel1cap1.Width)/2, (480-reel1cap1.Height)/2, color=$000000)
int = Interleave(bwp,reel1cap2)
est = int.depanestimate(range=1,pixaspect=1,zoommax=1,improve=false,trust=0)
dep = int.depaninterleave(est,pixaspect=1,prev=0,next=1,subpixel=2,mirror=0,blur=0)
acol = dep.SelectEvery(4, 1)
avgcap1cap2=overlay(reel1cap1,acol,mode="blend",opacity=0.5)

return(avgcap1cap2)

Man, this is so exciting.  I finally sat down & watched Puggo Grande in it's entirety & it was so great!  I can't wait to see this.

Since you asked, I for one would ask that you pay attention to the lightsabers, particularly in the duel shots.  We don't want another bubble-gum pink saber fiasco like in 2004 (not that I expect that would happen here anyway).

althor1138 said:

reel1cap2 = avisource("PATH TO FILE") #this is the video that will be altered.
reel1cap1 = avisourse("PATH TO FILE") #this is the video that will be the source that the other is matched to.

colw = reel1cap2.Width()
colh = reel1cap2.Height()
bww = reel1cap1.Width()
bwh = reel1cap1.Height()
borw = (bww-colw)/2
borh = (bwh-colh)/2
reel1cap2= reel1cap2.AddBorders((654-reel1cap2.Width)/2, (480-reel1cap2.Height)/2, (654-reel1cap2.Width)/2, (480-reel1cap2.Height)/2, color=$000000)
bwp = reel1cap1.AddBorders((654-reel1cap1.Width)/2, (480-reel1cap1.Height)/2, (654-reel1cap1.Width)/2, (480-reel1cap1.Height)/2, color=$000000)
int = Interleave(bwp,reel1cap2)
est = int.depanestimate(range=1,pixaspect=1,zoommax=1,improve=false,trust=0)
dep = int.depaninterleave(est,pixaspect=1,prev=0,next=1,subpixel=2,mirror=0,blur=0)
acol = dep.SelectEvery(4, 1)
avgcap1cap2=overlay(reel1cap1,acol,mode="blend",opacity=0.5)

return(avgcap1cap2)

 

That is so awesome, thanks a lot!  I will definitely be giving this a try.

canofhumdingers said:

Since you asked, I for one would ask that you pay attention to the lightsabers, particularly in the duel shots.  We don't want another bubble-gum pink saber fiasco like in 2004 (not that I expect that would happen here anyway).

Yeah, I'll be sure and keep an eye out for that as I do the color correction.

Puggo - Jar Jar's Yoda said:

That is so awesome, thanks a lot!  I will definitely be giving this a try.

Cool. Please let me know if it works.  I know of one more step to add if it doesn't but it requires a two pass so I thought simple is better to start with.

Thorr, it adjusts itself.  As long as the frames are identical in each capture.

I've changed the script so that it automatically will align all three captures of the first reel.  I've added a couple steps but I did a test run last night and this way is far superior to the old version(i'd say it works perfectly). It goes like this.

Step1:  Save this as avs file, input paths, and open it in virtualdub:

#reel1cap3=avisource("PATH TO FILE") #This video will be altered to match source video. Trim so that frames match source.
reel1cap2 = avisource("PATH TO FILE") #This video will be altered to match source video. Trim so that frames match source.
reel1cap1 = avisource("PATH TO FILE") #This is the source video.
blank=reel1cap1.blankclip() #blankclip used for forced scenechange in next line
inter=interleave(reel1cap2,reel1cap1,blank) #This will force a scene change after every comparison of one set of frames.
#inter=interleave(reel1cap2,reel1cap1,blank,reel1cap3,reel1cap1,blank) #This will force a scene change after every comparison of one set of frames.
return(inter)

Step2:

Add the Deshaker filter and configure pass 1.  The only changes I made was this: Scale:FULL,UsePixels:ALL,DeepAnalysis:30%.

Take note of the destination for the log file and press ok. Then you can press F5 to preview output from start.  Let the whole thing finish and then close Vdub and move on to next step.

Step 3:

Save this script to an avs file and input path to files and path to deshaker log. The output will be the average of all 3 captures or you can output each capture individually.

#cap3=avisource("PATH TO FILE") #this video will be altered to match source video. Trim so that frames match source.
cap2=avisource("PATH TO FILE") #this video will be altered to match source video. Trim so that frames match source.
cap1=avisource("PATH TO FILE") #this is the source video.
blank=cap1.blankclip()
#reel1cap3=interleave(cap2,cap1,blank,cap3,cap1,blank).depan(subpixel=2,pixaspect=1.0,inputlog="**PATH TO DESHAKER.LOG**").selectevery(6,3)
#reel1cap2 = interleave(cap2,cap1,blank,cap3,cap1,blank).depan(subpixel=2,pixaspect=1.0,inputlog="**PATH TO DESHAKER.LOG**").selectevery(6,0)
reel1cap2=interleave(cap2,cap1,blank).depan(subpixel=2,pixaspect=1.0,inputlog="**PATH TO DESHAKER.LOG**").selectevery(3,0)
reel1cap1 = cap1

#cap3w=reel1cap3.Width()
#cap3h=reel1cap3.Height()
cap2w = reel1cap2.Width()
cap2h = reel1cap2.Height()
cap1w = reel1cap1.Width()
cap1h = reel1cap1.Height()
borw = (cap1w-cap2w)/2
borh = (cap1h-cap2h)/2
#reel1cap3=reel1cap3.AddBorders((654-reel1cap3.Width)/2, (480-reel1cap3.Height)/2, (654-reel1cap3.Width)/2, (480-reel1cap3.Height)/2, color=$000000)
reel1cap2= reel1cap2.AddBorders((654-reel1cap2.Width)/2, (480-reel1cap2.Height)/2, (654-reel1cap2.Width)/2, (480-reel1cap2.Height)/2, color=$000000)
cap1p = reel1cap1.AddBorders((654-reel1cap1.Width)/2, (480-reel1cap1.Height)/2, (654-reel1cap1.Width)/2, (480-reel1cap1.Height)/2, color=$000000)
int = Interleave(cap1p,reel1cap2)
#int1=Interleave(cap1p,reel1cap3)
est = int.depanestimate(range=1,pixaspect=1,zoommax=1.5,improve=false,trust=0)
#est1=int1.depanestimate(range=1,pixaspect=1,zoommax=1.5,improve=false,trust=0)
dep = int.depaninterleave(est,pixaspect=1,prev=0,next=1,subpixel=2,mirror=0,blur=0)
#dep1=int1.depaninterleave(est1,pixaspect=1,prev=0,next=1,subpixel=2,mirror=0,blur=0)
cap1out=reel1cap1
cap2out = dep.SelectEvery(4, 1)
#cap3out=dep1.SelectEvery(4,1)
#avgcap2cap3=overlay(cap2out,cap3out,mode="blend",opacity=0.5)
#avgcap1cap2cap3=overlay(avgcap2cap3,reel1cap1,mode="blend",opacity=0.33)
maskcap2=mask(cap2out.converttorgb32(),cap2out.converttorgb32()).showalpha().invert()
avgcap1cap2=overlay(cap1out,cap2out,mask=maskcap2.showalpha(),mode="blend",opacity=1,greymask=true)

#return(avgcap1cap2cap3)
return(avgcap1cap2)

Ok, I just finished trying the earlier script (the first, simpler one).  I had to convert to YV12, then back to RGB, and also swapUV (my first attempt came out with a blue crawl!).  It worked better than I thought it would.  I'm not sure yet if I like the results - I have to study them carefully and see if they are doing what I want in the details.

Next I'll try your second set of scripts.

After I get some example clips, I'll post them.

Oh, althor, could you modify your second script for just two source files?  I think if I'm going to do this, I'm going to just choose the best matching pair of source files.  The alternate files are just for extreme scenes.

-- whoops, never mind -- was reading the file sizes wrong :P

How's it looking?

And by that, I mean can we take a peek? *hint hint* ;)

Puggo - Jar Jar's Yoda said:

Next I'll try your second set of scripts.

After I get some example clips, I'll post them.

SWEET!

Ok I've edited the script in post 139 to only align two captures.  I've commented out all the variables for a 3rd capture so if you want the capability, it is still there.

Awesome, thanks!!

I'm putting together a demo .mkv clip of the first 8 minutes, splicing the various pieces from the best two captures, border added, and resized to 1280.  I tried it last night with xvid4psp as suggested earlier in the thread, and I agree it seems to be a very nice tool for the job.

I'll also keep trying out althor's scripts.  The trouble with it, is that if the brighter clip blows out some of the whites, the script seems to also be blown out.  The "scripted" version may ultimately become another take from which to utilize in certain scenes.

10-minute .mkv sample posted, pieced together from two captures:

http://rapidshare.com/files/452924259/demo.mkv

It's 1280x480, so you'll need VLC or similar to play it.  I've noticed that if you play it in VLC, and tell it to switch to fullscreen, that it retains the aspect ratio and adds black bars, so that's the best way I know for watching it.

Some minor color correction is all, and the sound isn't perfectly sync-ed yet.
It's a rough, early sample, but you can hopefully see how dirty the film is at the start, but that it gets somewhat better as it goes along.

Just quickly scanned through the preview, Looks great, some parts are stunning, like the wide-establishing shot of luke on taun taun crossing the snow; alot more detail than i can remember seeing in other versions, and the colors truly are beautiful, especially in the wampa cave, that part looks great! Thank's for posting it :)

I wonder if maybe the wear on the top part of the image during the crawl could have been caused by the turning of the reel on the projector in some way as it comes and goes in..um... a timely way...

Looks great. It looks way better then Puggo Grande. I wonder if that is due to a better quality of the print or the scanning process being improved?

The colours need a bit more correction but it looks realy nice for what it is.