g-force

Moth3r said:

And the point g-force is making is that the GOUT stabilization script already includes upsizing as part of the EEDI2 anti-aliasing stage - the results are then downsized to 16:9 DVD resolution. If you want 1080p output it would make more sense to remove this downsizing stage and amend the Spline16Resize lines to produce your final desired output resolution.

Well stated. Thanks for your calm, rational explanations Moth3r.

-G

Okay, I'm going to eat crow now.

GradFun2DBmod shows some significant improvement in banding. I'm having difficulty using it at the same time as my script, due to the fact that the new pre-release of removegrain doesn't include clense for some reason, which I rely upon heavily in my script. I will need to find a work around.

Expect a new version of the script sometime in the future.

-G

How about keep Jar Jar and kill everything else about TPM?

-G

I just had to weigh in.

DJ, these screen shots look like oil paintings with halos everywhere. Why not just change the final resize in my script to whatever you want, instead of running it through additional filters that preserve very little detail, and oversharpen everything and then oversharpen everything again?!

Not my cup of tea.

-G

Orinoco_Womble said:   

Nice job G-force.

One question though, can someone please tell me how/where the GradFun2DBmod filter gets squeezed into the script?  Sorry for showing my ignorance, but I'm pretty new to Avisynth etc.

Thanks man! You could just stick the GradFun filter at the end, right before the subtitle stage, but I'm not convinced that it does anything beneficial yet. I'll try it out when I get a chance.

-G

I'm drooling. Please don't keep us without screen caps too long Arnie.d!

-G

Doctor M said:

fft3dfilter is awesome.  I probably couldn't live without it.  Still comparisons of before and after show it really does its thing for removing noise.

What I don't understand about it is that in motion it makes it look like the camera lens is dirty.  Besides softening the picture, you get a soft pattern that looks a bit like film grain but doesn't change pattern from frame to frame.

 Does anyone know what I'm talking about?

It is a handy filter, but I have noticed the artifact that you mention. I had this problem with many of the earlier scripts. The trick is to not do too much FFT filtering, which is why the script uses the FFT filter only to determine the motion vectors, and then as a median between the mo-comped and original pixels. One way to look at this is, it only uses the fft filter if the fft is inbetween the mo-comped pixel and the original pixel.

I think the reason there is this left-over haze of noise that persists frame to frame if you're not careful is because of the MPEG compression of the original DVD. The compression tries to resist change, hence it takes an otherwise random noise and only changes as little of it as possible to make it still look like random noise. Then when you apply the filter, the noise left over is just the persistant noise that the original DVD compression decided to make less random.

-G

DJ,

just to let you know, I'm very excited about this project. I would love to help, but I have to admit that it can only be in a limited capacity, as I have a newborn now, and I just can't devote the kind of time I used to. I think the first step would to be to get the best capture you can, and do the best IVTC job you can, then you'll have something I can help with.

-G

Looks great to me DJ!

-G

Moth3r said:

In line 60, should the "Nop()" be replaced with a "last"?

 

Indeed, it should. Oops! I'll update the script on the first page.

Thanks Moth3r!

Doctor M said:

Noise reduction: There is some serious degrain/denoise/fft3dfiltering going on.  It seems like you are guilty of scrubbing all the film grain with the noise. 

Guilty. This was necessary in order to remove shake that was the result of poor telecine and then the DVD compression caused the shake to move different parts of the frames differently.

I'm assuming you all are being careful not to lose too much detail with the noise. 

Yes, I actually worked on this for the better part of a year.

Interlace artifacts: When the GOUTs first came out and the first scripts for clean up were being discussed, filters for the interlace artifacts were heatedly debated.  I don't see anything being used on them here.  (Or did I miss it.)

You missed it (them). It actually takes the best of 2 different anti-aliasing methods.

Sharpening: Different sharpening algorithms can improve different features of an image.  For any given filter you can only go so far before ringing or edge enhancement artifacts become an issue.  I've found that MSharpen used mildly in conjunction with fft3dfilter's sharpen provides a nice balance.  Here's some other sharpen filter ideas: http://www.aquilinestudios.org/avsfilters/sharpeners.html

I spent the year optomizing this as well.

Banding: Fft3dfilter use can result in color banding or posterization.  I'm definitely seeing some of that in the stills and clips (look at the close up of Han's hand).  The best answer is to either back off of fft3dfilter or use a deband filter. http://avisynth.org/mediawiki/External_filters#Debanding

I think the banding is coming from Dark Jedi's encoding. It does look bad on those stills.

Sample clip: In the still below, the silhouetted figure walking in the background leaves trails.  Trippy and an artifact of agressive temporal noise filtering.

From good old DVNR smear on the GOUT.

-G

well, it's late, but I thought I would add some thoughts. DJ, those shots look all wrong to me. I don't have anything to comapre with right now, but they seem overly bright, overly saturated, and there are too many sharpness artifacts.

dark_jedi said:

I want to get these as perfect as I can so any changes have to be done today\tonight so I can start it all tomorrow night.

oops, forgot to mention in my PM that there is no more recent version than what is on page 1 of this thread.

Good luck!

DarthJardus said:

Also, in HC, it still says 4:3 and not 16:9.   Am I ok since I used the script on page 1 or was I supposed to change that too?

It's been a while, but I think you have to choose 16:9 in HCenc for the aspect ratio to come out right. I highly recommend encoding a small portion first before attempting a full encode. There's just too many things that can go wrong.

pupil said:

G-force, just checking, but are you planning on developing this script any further, especially for ESB and RotJ? Not that I think it's necessary, it's just that I'd rather know before doing 2 more 4day encode-athons for these 2 movies :)

ANH came out perfectly btw, watched it the other night and it looks stunning compared to the GOUT DVD. I'm ripping the selectable Greedo subs off the DVD tonight to add to my mkv file and then it's perfect!

I wish I could encode in 4 days, more like 4 weeks for me.

No, I don't have any plans to develop this further. I've taken it as far as I can.

-G

INv8r_ZIM said:

########## subtitle stage
episode==4 ? ANH(last,PAL) : episode==6 ? ROTJ(last,PAL) : Nop()

I know it works for me. The No-op at the end catches everything that is not 4 or 6. I'm stumped.

-G

dark_jedi said:

"EEDI2"

ooh! that means you haven't seen the latest anti-aliasing! Looks better than just NNEDI by itself.

dark_jedi said:

"Trim(70600,72000)"

this will do the trick.

Yep, that'll do it.

-G

dark_jedi said:

it says this,

Avisynth open failure:
FFT3DFilter: bt must be -1(sharpen), 0(kalman), 1,2,3,4(wiener)
(D:\Star Wars GForce.avs, line 26)

so it is line 26 of the script,I can't see what is wrong.

Thanks for sharing your time.

 

I think I changed bt since you tried the script last, and now you need a newer version of FFT3DFilter.

according to the FFT3DFilter version history:

Version 1.9.2 - 11 september 2006 - added mode bt=5

so anything later than 1.9.2 should work.

-G

pupil said:

But for my purposes, I want square pixels as I will be playing this on my PC monitor. I've finished compressing the video, added the flag to make it 853x368 (56 pixels from the top and bottom of the image cropping the black bars as much as possible) in the mkv and it has come out at exactly the same resolution as the original DVD.

 

Wow, all this 4:3 anamorphic non-square pixels to 16x9 square pixels gets confusing. Glad you figured it out!

-G

dark_jedi said:

well here is the error I am getting now,does anyone know what this means?

Oh no, I can't see the picture on my work PC, because my company blocks so much content. Can you tell me what it says?

-G

satanika said:

g-force said:

Spline16Resize(960,480) #for NTSC

Spline16Resize(960,576) #for PAL

Wouldn't it be something like 854x480 & 1024x576??

 

That would be the case for square pixels, but the pixels aren't square.

-G

pupil said:

Basically, after processing the d2v through this avs script, I am left with an image that isn't anamorphic and has the black borders around the image.

The output of the script is truely anamorphic (that's why everyone looks thin and streched if your player plays it back as a 4x3 image and doesn't apply the anamorphic filter). The black bars are due to the fact that SW is not in 16x9 aspect ratio, it's actually even "wider", so some black bars are required on the top and bottom to fill the rest of a 16x9 field. Probably your best bet would be to add another resize at the end of the script instead of your crop to do what a DVD player does when is sees the anamorphic flag.

Spline16Resize(960,480) #for NTSC

Spline16Resize(960,576) #for PAL

-G

INv8r_ZIM said:

Anyone suggest a decent way of doing the 3:2 pulldown?  I've been chugging along fine with ESB GOUT and a number of other projects in CCE and HC, but for some reason when it comes to ANH GOUT the mpeg always ends up terribly jerky, even though the uncompressed avi is fine.  Turning off pulldown from a run through HC fixed it, but obviously it'll have to be done at some point to be DVD compliant.

Hmm... I don't have any experience with encoding from an avi, due to my increasingly limited hard drive space. When I encode directly from the avisynth script, and choose "NTSC" and "Make DVD Compliant" in HCencoder, everything turns out fine. I wonder if avisynth is sending more information to HCencoder than the avi does? Just a thought. Hopefully someone more knowledgable will chime in.

-G

Knightmessenger said:

Could you show us an example of these lines. 

 check out the pics that INv8r_ZIM posted on the last page. If you look near the borders you will see what I'm talking about.

-G

INv8r_ZIM said:

Tried the new script, but am getting a fairly unhelpful response from AVISynth ("The Script's Return Value Was Not a Video Clip")

Anyone else having this problem? Might be a problem when you copied it out of the page and into your text file.

-G