another restore function

[MOmonster]

Bad telecined sources with blended fields seems to be normal in my country. The possible solutions for this problems are really rare. For example tdeint(mode=1, tryweave=true) with unblend and a decimater or the the great but really complex function restore24.
That's why I have written the function srestore (18.11.2009).
You can download the necessary plugin, Masktools v2 (latest version), here.

For more informations please read the readme or take a look at the wiki.
I hope this is a helpful function. Nice tryout.

[scharfis_brain]

MoMonster, I didn't test your unblender due te lacking time

So I have one question: how does it handle the Starship Voyager Intro video?
Is it able to deblend the nebula scene without jerkyness?

[Didée]

Originally Posted by scharfis_brain

how does it handle the Starship Voyager Intro video?
Is it able to deblend the nebula scene without jerkyness?

As far as I tried, it fails big time on this clip. But it must be admitted that this intro is very difficult indeed, and probably I didn't know how to tweak CDeint's setup correctly.

(MOmonster - if you're still interested in Voyager's intro, drop me a PM.)

What I noticed with the latest CDeintMod: it seems to mistreat chroma planes when it is fed with YV12 sources, as CDeint's output shows chroma interlacing, then. When converting to YUY2 beforehand, it's clean.

[MOmonster]

Also Cdeintmod only choose the right frame of the bobbed input, so I can´t understand this chroma problems. I´ll take a closer look on the next version.
I don´t any longer recommed fdecimate after cdeint. It has a strange behaviour in some scenes. Decimate works fine. I don´t tested Tdecimate.
I just finished a newer and better version, that give me sometimes better results than restore24, but I have to make some more tests with this version.
I´ll send you my emailadress soon.

[MOmonster]

@all
Once more I updated the modded version (not the silent - will come soon).
There is many new in this version. Now it works much more fluid, also for the Starship Voyager Intro video. It has some other parameters and runs more stable than the last both versions. Like the first you can also use it together with fdecimate, but it is recommed to use decimate.
For more informations please read my updated second post.

[yaz]

@momonster
thx for the improvements! i'll try to give it a go at the weekend. (i'm still fighting w/the same source )

thx
y

[Didée]

I gave it a quick try on CDeintMod yesterday. Seems to work pretty good, overall.

While the blend removal was quite successful (perhaps even better than R24 - I didn't count the misses for both), it produced yet too much skipped frames on scenes where motion happens only in small areas.

Though I don't get fully get the mechanism behind it, the function shows good potential for sure. Good work, Momonster!

Note: Work on R24 resumed

[MOmonster]

I work on a better condition function for Cdeintmod, that should save more clean frames, but ofcourse it would never save the clear frames as good as restore24, because it just take the bobbed input and choose one of two frames to output. That mean that the bobber you choose give two frames for one original frame. My function just try to find out what is the blend and give back the clear field. For my sources, everytime when tfm is not able to find a match it´s because a blend, that´s why this simple way works that way (till now smoother and better than the mod version, but there will be further tweaking).
The mod version I created to save more clear fields. So it looks to some more conditions, but there is still some work necessary.
@Didee
Ok, I try to explain short, how the blenddetection work.
The condition is that the blend is the product of the two clear fields around. The blenddetection use a smart bobbed clip, this makes the test easier. I found out, if the blendlevel (the level, the clear fields around are merged together) is 50%, the differences between the blending to the field before and the field after are nearly the same. If the blendlevel is 20%, the difference to the field after is smaller than the difference to the field before.
And this seems logical. So my function try to create the blend with the both differences to the clear fields around. Then it compares the created blend with the original field. If they match, we know, this is a blend.
To make this process more accurate, the function compares the result of the first blendtest with the results of the next blendtest.
I also tried to make it more accurate by using the total difference, but it only slows down things a little bit, and don´t really help, the lumadifference seems to work quite well. In your restore24 thread I posted my first basic function, maybe this is easier to understand.
If there are more questions, just ask me.
@scharfis_brain
If you could upload a small part of the intro, I could work a little bit with that source to tweak my function.

[yaz]

@momonster (and other gurus interesting)
i've uploaded two short examples. link and comments are here. sorry for crosslinking but i think the files belong there even if we discuss the problems here (hopefully). and i don't want to crosspost.

@momonster
'man proposes God disposes' ... of course, i had time far less than i expected but i'd made some further short tests w/the modded versions. i must confess, i found no difference betw them. however, the results were amazing. most of the blended frames were recovered or cleared significantly. what remained is that strange ghosting i outlined in the other thread.
further, i felt as if i had lost 'too much' clear frames while decoming was defective on frames where it'd been needed only on small regions. however, i haven't checked these more throughly but, imho, they are not 'fatal', so, i could live w/them happy.

a short question. you, and other gurus too, recommend to restore fps24 ? why do u think it's beneficial ? i see that wout some decimation the stream gets jerky but i want to keep fps25. any ide how to do so ? (other than fps25 is hella jerky on a pal tv)

thx a lot
y

[Mug Funky]

field-blends use up buttloads of bitrate and require bobbing on playback... that's the main reason to restore the non-blended source. also if you plan on standards-converting it a second time and don't want to have blurry crap as an output.

if you have bitrate to spare, you can easily encode an interlaced xvid and play back with on-the-fly bobbing. it's quite a good solution for anime, where there's lots of static scenes.

[MOmonster]

@yaz
Ok, I dowloaded your clips. What a strange conversation, a soft picture, but no analoge noise and not to heavy artefacts. I´ll take a closer look, if I have a little bit more time.
Why decimating to 24fps?
In most cases the original has 24fps and than converted with strange conversations like temporal fieldshifting fielddoubling and of course fieldblending to 25 or 29,976fps. You see it shouln´t be 24fps, but 23,976fps. Sometimes the original has only 17fps and one time I have a source, that was already speed upped to 25fps before the strange conversation. So we can´t say that it is general right to decimate the clip to 24fps. So you have to decide, if the 25fps output looks already smooth enough for you, there is no need to decimate. If the 24fps decimated clip have a smoother motion for you, than you can decimate and than speed up the source to 25fps, that you have your pal standard.

As I said before the modded version works not so good till now. I already have another condition function for Cdeintmod. It works better than now and with the right settings already better than the silent Cdeint with tfm. I will update the function, if I have a little bit more time, till now it´s recommed to use the non modded version.

I also think about another modified version that is not only able to use one of two, but is able to use one of three frames to save more clear fields, but I have no such strange sources. We´ll see, what brings us the future.

[Didée]

Screenshot of those "dark stripes", please (can't get yaz' samples, currently) ?

Some time ago, scharfis_brain had a source with something he called "negative blending", and this here sounds similar.
For scharfi's sample, I fiddled a spatio-temporal repair function that seemed to work sufficiently. Perhaps it can be applied on yaz' source, too.

[yaz]

Originally Posted by Didée

Screenshot of those "dark stripes", please

here you are

Originally Posted by Didée

(can't get yaz' samples, currently) ?

naah ... what to say ...
thx
y

[MOmonster]

I know the strong limitations of this server. I uploaded the clip on another server. Download it here:
http://www.megaupload.com/?d=22K8DY2Z

[vcmohan]

When I click on the link it takes to megaupload.com. What to do next to get the image?

[MOmonster]

There is a small button "please wait ... seconds". Wait the time and then click on the link.
Maybe you need the macromedia flash plugin to see the button. If you don´t want to install just try the link of yaz. I don´t want to upload the clip once more on another server.

[outcomes]

^

Got it

[cretindesalpes]

Hello,

I have a telecined video with a weird kind of progressive blending between frames (I suppose it is intended to smooth transitions when displayed on cathodic screens). Is it possible to restore the original frames, and how ?

[malze]

I was wandering if there is a 64-bit verison of Srestore plugin or similar plugin of this.

[Didée]

MOmonster - I could only give a quick try on yesterday's preview version. Overall fluidity was indeed better than before, you're on a good way.
Still, there were noticeable problems when motion appears only in small areas of the frame. In a scene where nothing moved but the mouth of the actors speaking, it reminded somewhat of an animation @ 8fps or 12fps - the mouths were "hacking". However, small motion is always problematic. Have been fighting with that long enough


@ yaz

A solution I don't have for you. But a strategy. One would have to

1) make a basic restoration by either CDeintMod or R24. After this, most blends have gone. Theoretically, in ideal case, all of the remaining blends are in the places where the good original fields are missing in the source already.
(Do I remember correctly that the good fields are always missing where two consecutive blends are occuring?)

2) Run a 2nd pass, where all blends are replaced with the previous frame. This turns the remaining blends into dup's.

3) run a 3rd pass, with a routine that replaces duplicates with motion compensated frames, computed from N+1 and M-1.


Step 1 is obvious, and both Restoring functions should be able to deal with it.

Step 2 is the problematic one, of course. R24 can not do that, but CDeint should, or should be possible to be changed so that it can.

Step 3' s function does not yet exist AFAIK, but wouldn't be that hard to do. (I wonder why nobody did it, yet)