althor1138

With a little bit of luck I am finished with posting the raw captures. The JSC trilogy,'97SE trilogy, TPM, and the 82 P&S are all finished.  I will be keeping the archives for at least a couple of months in case anybody experiences trouble getting them. SW'97se and ESB'97se have been re-upped and it says so in the title so don't get confused and try to get these 2 unless it says REUP.  

The JSC trilogy currently has ivtc scripts for ESB and ROTJ which Mallwalker has generously provided. I was going to start on SW but I believe it is better left to him.  His ivtc scripts come with all sorts of awesome stuff like reel info, gout frames,etc. 

The '97SE trilogy all have ivtc scripts that accompany the raw files. These are complete by my standards.  For some reason I can't recall, I included the laserdisc and thx stuff at the beginning.

TPM also has an IVTC script included in the archive.

I've tried to IVTC the 82  P&S but the cadence is crazy and I can't really figure it out.

Does anybody have any plans to do a JSC release derived from the raw caps?

davros said:

althor1138 said:

I'm getting myself one for christmas.  I have about 2-3 minutes of sw on 35mm that I want to scan.  I have a galaxy s2+ and a program that can take png's.  If it works I'll post some shots.

Sounds good! What film ? Please post any examples

 It is Star Wars on Eastman I think.  I will definitely post some pictures when I get it. What type of film is yours printed on, if you don't mind me asking?

Thanks for sharing your scripts! This is great for me.  I had not even approached ROTJ yet.

BTW, I have uploaded,downloaded, and confirmed TPM-RAW so anybody wanting to get that should be able to.  Mileage might vary but in any case I now have enough HDD space to keep all of the archives just in case somebody needs some parts.  Also, Jbinup seems much more stable and modern than powerpost imo.

The 82P&S raw files are next up in the chute.  I have already created all of the files and will probably make them available in a few days. This is the only capture I have that I captured my own sound for.  Before anybody decides they want it they should understand that there is a lot of noise on side 1, both video and audio. Side 2 seems rather ok for such an old release though.

I'm getting myself one for christmas.  I have about 2-3 minutes of sw on 35mm that I want to scan.  I have a galaxy s2+ and a program that can take png's.  If it works I'll post some shots.

We've been able to successfully download and extract ROTJ'97.  I just need to re-upload about 5 gb worth of rar's and some new par's and it will be possible to extract the files.  There is no hope for SW/ESB'97 so those will get an entirely new upload.  There's been some bad storms here for a few days and the electricity is going out randomly so I'm going to wait until it stabilizes before attempting to post anything.

I'd definitely recommend holding off.  SW/ESB 97 are going to be re-upped for sure.  Don't even mess with the existing files on Usenet.  I have switched to a fiber connection now so that should be more consistent I hope.  I'm currently trying to see if I can get Chewtobacca up and running with ROTJ97.  If that happens then I'll simply be able to upload what's missing/broken.  In any case the next thing I upload I'm going to download myself and test it so I'd wait until I give the green light if I were you.  Thanks for all the patience guys.

Hey Andrea, your pm box is full.  I'll just post it here though.  It should work with ntsc and pal without any extra input from the user.

EDIT:  Had to separate luma and chroma filtering.  The chroma sampling was actually sampling luma and removing too much noise.  Separating luma and chroma filtering has caused it to run very slow but the output is much better quality.  So it is.

function LDnoise(clip input, float "strength", int "mc", int "temporalframes",\

int "blksize", int "search", int "searchparam", int "overlap", int "dct")

{

 

# Set default options. 

 

width=width(input)

height=height(input)

strength=default(strength,1)

temporalframes=default(temporalframes,1)

mc = default(mc, 0)

 

# Prepare supersampled clip.

 

super = input.MSuper(levels=6,chroma=true)

 

# Motion vector search.

 

b5vec = MAnalyse(super, delta=5, isb=true,chroma=true, search=search, searchparam=searchparam, overlap=overlap, dct=dct)

b4vec = MAnalyse(super, delta=4, isb=true,chroma=true, search=search, searchparam=searchparam, overlap=overlap, blksize=blksize, dct=dct)

b3vec = MAnalyse(super, delta=3, isb=true,chroma=true, search=search, searchparam=searchparam, overlap=overlap, blksize=blksize, dct=dct)

b2vec = MAnalyse(super, delta=2, isb=true,chroma=true, search=search, searchparam=searchparam, overlap=overlap, blksize=blksize, dct=dct)

b1vec = MAnalyse(super, delta=1, isb=true,chroma=true, search=search, searchparam=searchparam, overlap=overlap, blksize=blksize, dct=dct)

f1vec = MAnalyse(super, delta=1, chroma=true, search=search, searchparam=searchparam, overlap=overlap, blksize=blksize, dct=dct)

f2vec = MAnalyse(super, delta=2, chroma=true, search=search, searchparam=searchparam, overlap=overlap, blksize=blksize, dct=dct)

f3vec = MAnalyse(super, delta=3, chroma=true, search=search, searchparam=searchparam, overlap=overlap, blksize=blksize, dct=dct)

f4vec = MAnalyse(super, delta=4, chroma=true, search=search, searchparam=searchparam, overlap=overlap, blksize=blksize, dct=dct)

f5vec = MAnalyse(super, delta=5, chroma=true, search=search, searchparam=searchparam, overlap=overlap, blksize=blksize, dct=dct)

 

# Motion Compensation.

 

b5clip = MCompensate(input,super, b5vec)

b4clip = MCompensate(input,super, b4vec)

b3clip = MCompensate(input,super, b3vec)

b2clip = MCompensate(input,super, b2vec)

b1clip = MCompensate(input,super, b1vec)

f1clip = MCompensate(input,super, f1vec)

f2clip = MCompensate(input,super, f2vec)

f3clip = MCompensate(input,super, f3vec)

f4clip = MCompensate(input,super, f4vec)

f5clip = MCompensate(input,super, f5vec)

 

# Create compensated clip.

 

interleaved = mc >= 5 ? Interleave(f5clip, f4clip, f3clip, f2clip, f1clip, input, b1clip, b2clip, b3clip, b4clip, b5clip) :

\ mc == 4 ? Interleave(f4clip, f3clip, f2clip, f1clip, input, b1clip, b2clip, b3clip, b4clip) :

\ mc == 3 ? Interleave(f3clip, f2clip, f1clip, input, b1clip, b2clip, b3clip) :

\ mc == 2 ? Interleave(f2clip, f1clip, input, b1clip, b2clip) :

\ mc == 1 ? Interleave(f1clip, input, b1clip):

\ input

 

#Perform DFTTEST

 

yparams="""dfttest(y=true,u=false,v=false,f0beta=0.5,sigma=0,dither=0,sbsize=25,sosize=20,tbsize="""+string(temporalframes)+""",tosize="""+string(temporalframes)+"""/3,nstring=\

    "a:"""+string(strength)+""" \

    "+string(current_frame)+",0,"""+string(height-40)+""",10 \

    "+string(current_frame)+",0,10,10 \

    "+string(current_frame)+",0,10,"""+string(width/4)+""" \

    "+string(current_frame)+",0,10,"""+string(width/2)+""" \

    "+string(current_frame)+",0,10,"""+string(width/1.5)+""" \

    "+string(current_frame)+",0,10,"""+string(width-40)+""" \

    "+string(current_frame)+",0,"""+string(height-40)+""","""+string(width-40)+""" ")

    """

 

 

 yfilter=eval("scriptclip(interleaved,yparams)")

 

 uvparams="""dfttest(y=false,u=true,v=true,f0beta=0.5,sigma=0,dither=0,sbsize=25,sosize=20,tbsize="""+string(temporalframes)+""",tosize="""+string(temporalframes)+"""/3,nstring=\

    "a:"""+string(strength)+""" \

    "+string(current_frame)+",1,"""+string((height/2)-30)+""",3 \

    "+string(current_frame)+",1,3,3 \

    "+string(current_frame)+",1,3,"""+string((width/2)/2)+""" \

    "+string(current_frame)+",1,3,"""+string((width/2)-30)+""" \

    "+string(current_frame)+",1,"""+string((height/2)-30)+""","""+string((width/2)-30)+""" \

    "+string(current_frame)+",2,"""+string((height/2)-30)+""",3 \

    "+string(current_frame)+",2,3,3 \

    "+string(current_frame)+",2,3,"""+string((width/2)/2)+""" \

    "+string(current_frame)+",2,3,"""+string((width/2)-30)+""" \

    "+string(current_frame)+",2,"""+string((height/2)-30)+""","""+string((width/2)-30)+""" ")

    """

 

uvfilter=eval("scriptclip(yfilter,uvparams)")

 

return SelectEvery(uvfilter, mc * 2 + 1,mc)

 

}

Ha. I didn't even think about making it PAL compatible.  That could easily be done.  How many lines are above and below the picture for a pal widescreen release?

Thanks. I've updated the first post to mirror your function.  Avspmod adds some text formatting which shouldn't affect functionality but obviously if you can't run the function then it should be laid out so that it will work for everyone.  If that's even possible with a program like avisynth lol.  I'm using avisynth 2.58 btw.

It's good that you had the knowledge to make it work :)

EDIT:  I also named it LDdenoise at first but then I thought dnoise sounds the same as denoise and there's one less character to type.  I'm a lazy bastard lol.

Will do.  Thanks.

I don't blame you at all.  I still have the par's and rar's for rotj97.  I am able to verify the files with the sfv and pars unlike sw97.  I'll provide any files that you request.  

On a related note, I believe I have discovered that it isn't my isp,usenet provider, or even user error.  It seems that the program I'm using for posting is buggy in that it causes crc errors.  I'm immediately switching back to the old program even though it posts slower.  The program is camelsystem powerpost btw.  I will be using powerpost 2000 from now on.

Here is something that is maybe a bit more informative.  The top is fft3dfilter.  Bottom is LDnoise. I adjusted the two images until I couldn't see any grain in the sky.  As best as my eyes can do anyway.  If you look close there is detail lost in the sand on the fft3d frame.

EDIT: Here's the original for comparison as well.

I chose this frame with some very high settings.  I think it shows how well it works in choosing noise over detail. The fine strands of hair stay pretty much intact after the denoising. Also, the detail in the soldier's uniform in the background seems to have weathered the eraser well too.

BTW, I've found that enabling temporal frames with this setup doesn't always do much.  The most important thing it seems is to have enough noise samples to get a complete "picture" of all the noise contained in the picture.  I think I could add at least 2 more noise sampling areas to the script without interfering with subtitle areas.

I came up with this idea a couple of weeks ago and it worked a lot better than I thought it would so I'm posting it here to hopefully get some feedback and so others can use it if they want.

There are a few requirements for this function to work properly:

1.) Clip needs a resolution of 720x480.

2.) Must be in YUY2 colorspace.

3.) Must be widescreen format. (This denoiser samples the black bars for noise and uses this information to denoise the film area.)

4.) Requires MVTOOLS2 and DFTTEST along with knowledge of avisynth.

function LDnoise(clip input, float "strength", int "mc", int "temporalframes",\

int "blksize", int "search", int "searchparam", int "overlap", int "dct")

{

 

# Set default options. 

 

strength=default(strength,1)

temporalframes=default(temporalframes,1)

mc = default(mc, 0)

 

# Prepare supersampled clip.

 

super = input.MSuper(levels=6,chroma=true)

 

# Motion vector search.

 

b5vec = MAnalyse(super, delta=5, isb=true,chroma=true, search=search, searchparam=searchparam, overlap=overlap, dct=dct)

b4vec = MAnalyse(super, delta=4, isb=true,chroma=true, search=search, searchparam=searchparam, overlap=overlap, blksize=blksize, dct=dct)

b3vec = MAnalyse(super, delta=3, isb=true,chroma=true, search=search, searchparam=searchparam, overlap=overlap, blksize=blksize, dct=dct)

b2vec = MAnalyse(super, delta=2, isb=true,chroma=true, search=search, searchparam=searchparam, overlap=overlap, blksize=blksize, dct=dct)

b1vec = MAnalyse(super, delta=1, isb=true,chroma=true, search=search, searchparam=searchparam, overlap=overlap, blksize=blksize, dct=dct)

f1vec = MAnalyse(super, delta=1, chroma=true, search=search, searchparam=searchparam, overlap=overlap, blksize=blksize, dct=dct)

f2vec = MAnalyse(super, delta=2, chroma=true, search=search, searchparam=searchparam, overlap=overlap, blksize=blksize, dct=dct)

f3vec = MAnalyse(super, delta=3, chroma=true, search=search, searchparam=searchparam, overlap=overlap, blksize=blksize, dct=dct)

f4vec = MAnalyse(super, delta=4, chroma=true, search=search, searchparam=searchparam, overlap=overlap, blksize=blksize, dct=dct)

f5vec = MAnalyse(super, delta=5, chroma=true, search=search, searchparam=searchparam, overlap=overlap, blksize=blksize, dct=dct)

 

# Motion Compensation.

 

b5clip = MCompensate(input,super, b5vec)

b4clip = MCompensate(input,super, b4vec)

b3clip = MCompensate(input,super, b3vec)

b2clip = MCompensate(input,super, b2vec)

b1clip = MCompensate(input,super, b1vec)

f1clip = MCompensate(input,super, f1vec)

f2clip = MCompensate(input,super, f2vec)

f3clip = MCompensate(input,super, f3vec)

f4clip = MCompensate(input,super, f4vec)

f5clip = MCompensate(input,super, f5vec)

 

# Create compensated clip.

 

interleaved = mc >= 5 ? Interleave(f5clip, f4clip, f3clip, f2clip, f1clip, input, b1clip, b2clip, b3clip, b4clip, b5clip) :

\ mc == 4 ? Interleave(f4clip, f3clip, f2clip, f1clip, input, b1clip, b2clip, b3clip, b4clip) :

\ mc == 3 ? Interleave(f3clip, f2clip, f1clip, input, b1clip, b2clip, b3clip) :

\ mc == 2 ? Interleave(f2clip, f1clip, input, b1clip, b2clip) :

\ mc == 1 ? Interleave(f1clip, input, b1clip):

\ input

 

#Perform DFTTEST

 

params="""dfttest(y=true,u=true,v=true,f0beta=0.5,sigma=0,dither=0,sbsize=25,sosize=20,tbsize="""+string(temporalframes)+""",tosize="""+string(temporalframes)+"""/3,nstring=\

    "a:"""+string(strength)+""" \

    "+string(current_frame)+",0,440,10 \

    "+string(current_frame)+",0,10,10 \

    "+string(current_frame)+",0,10,340 \

    "+string(current_frame)+",0,10,680 \

    "+string(current_frame)+",0,440,680 \

    "+string(current_frame)+",1,440/2,3 \

    "+string(current_frame)+",1,3,3 \

    "+string(current_frame)+",1,3,340/2 \

    "+string(current_frame)+",1,3,680/2 \

    "+string(current_frame)+",1,440/2,350 \

    "+string(current_frame)+",2,120,3 \

    "+string(current_frame)+",2,3,3 \

    "+string(current_frame)+",2,3,340/2 \

    "+string(current_frame)+",2,3,680/2 \

    "+string(current_frame)+",2,440/2,680/2")

    """

filter=eval("scriptclip(interleaved,params)")

return SelectEvery(filter, mc * 2 + 1,mc)

 

}

There are only 3 parameters that you need to adjust really.  They are:

Strength: It defaults to 1.  Valid range is .01 and up.  1-10 is a good area to stay within. Every source needs tweaked however, so numbers can vary wildly.  There is no set and forget.

Temporalframes: Defaults to 1. 1 enables spatial only filtering.  Must be an odd number. Numbers above 9 are overkill and will probably cause a crash. I recommend 1 or 3.

MC: Defaults to 0. Valid range is 0-5. Defines the number of motion compensated frames. I recommend setting this to 5 if you have enabled temporal filtering with the temporalframes parameter.

This is basically a motion compensated dfttest wrapped into a function.  It's different from just running dfttest and letting the denoiser blindly search for noise though.  It samples noise from 5 separate blocks in the widescreen bars, for the frame that it is processing, in the luma and chroma planes and applies this information when denoising. If you enable temporal frames it starts comparing noise from the previous and future frames as well.  The motion compensation prevents stuff like laser bolts from getting erased when temporal filtering is enabled, etc.  It has the same problems as most denoisers and if you crank it up everything will turn to plastic.  It's strong point is that it is much better at keeping detail with less artifacting. It can cause banding with higher values, especially in the black bar areas.

This won't work on the GOUT,for example, because there isn't any analog noise in the black bar areas as far as I know. BTW, this will work with the JSC captures because I've set the coordinates for the noise sampling blocks outside of any areas that would contain subtitling.

The LOLZ are unnecessary but your inevitable banning definitely won't be.

I have a 4TB HDD coming in the mail.  This will allow me to keep the files for everything without having to delete them before moving on to the next raw capture post. So, maybe by next week I can start again with sw97. 

Colton, I'm not a sound guy, so any advice I have for you would be ill advice, lol.

Also, it would be nice to have confirmation of ESB and ROTJ.  I don't believe they will be problematic but still nice to know what I need to fix before moving on.  

Here is a link to a 1.4gb encode(no sound) of sw97. The link expires after a week or two btw.  It's raw other than the ivtc and compression. I'll still make the raw files available but I thought this would be a good way to check it out with a much shorter download time.

Well, it seems nothing will work to recover the files.  I'll re-up SW97 when I can.  Is there a way to remove a post from the newsgroup?

Here is a temporary link that will last for 7 days.  I had to make new rar's and then generate par's like the last time so let's hope it works.  The par's are about 850mb.

I've been using 64bit multipar to make the par's.

I'm assuming you mean for Star Wars '97.  I must be doing something wrong, but I always check everything twice before uploading.  I'll either upload some more par's or the parts that you need.  It might be later though because I have some things to do atm.

The 1997 SE Trilogy is now available.

I have 2 more raw captures to post:

1982 Pan and Scan

The Phantom Menace 

I'll probably start working on doing that next month.  After that, I plan on re-doing the IVTC scripts for the JSC and hopefully can learn from my mistakes last time and use all of the tips given to me to fix all of the previous problems.

Yup. I'm in Sweden.  I was holding out on selling my player because I was supposed to do some capturing for someone but that has never materialized. So, I'm actually going to sell my player now.

 I have found that combining captures doesn't have much to offer anyway. I have posted all of my raw captures though, so if you wanted to capture yours and have a crack at it my captures are freely available to experiment with.

coltonlolz said:

You_Too, I also checked with Google Translate, and it said "The Empire Strikes Back" when I put in Det Imperiet Slår Tillbaka in. You are the silly one in that paragraph, :)

I'm not Swedish but I can speak the language and I promise you it's Rymdimperiet Slår Tillbaka.  The fact that you are disputing this baffles me when it's on the Swedish movie poster and several Swedes have told you already. It's insulting to be perfectly honest.

Atm I am solely focused on making the raw captures available. In that respect, I am nearly finished.

I've never owned one, but from what I understand, the cld-s104 offers a clean composite signal and they are rather cheap.

It was kind of a base model back then but nowadays less filtering is desirable if you are going to capture with a device that has better filtering capabilities anyway.