Despot and Descratch combination for 8mm film transfer

[mammo1789]

Hi guys,

I tired to transfer some 8mm film footage on to the pc. I put the projector near the white paper on wall( my old laptop was used as wall) and on top i put the camera as near as to the projector, and video out from the camera was inputted in to tv card and i did capture in lagarith 720 x576 ( i know many use the camera to record on the tape or memory card but i think this is better approach to use only the lens of the camera and to record on pc in what ever format you want)
I tried the video fred 8mm script although looks impressive on his footage ( it is scanned by the way not recorded with camera like me and it blurs and sharpens very much also I think it as many necessarily steps, and depan is not so good as deshacker so i made very simple script with despot descratch and deflicker but even if descratch has some impact ( very little i tried with mindf 1 instead of the default 2 and it did improve little. The despot is not doing anything at all i tried couple of combination and it still doesn't remove the spots which are very big by the way ( you cant miss them )

Any thought suggestion i uploaded the video pictures of the setup and file

the scipt
AVISource("D:\ggggggg.avi")
FlipVertical()
SeparateFields()
ConverttoYV12
Deflicker()
#Crop(0,0,720,288) #to show test
#i = last
# Compare half-frames with and without noise reduction
DeSpot(p1=35, p2=14, pwidth=70, pheight=70, mthres=25, mwidth=20, mheight=15, interlaced=false,
\ merode=33, ranked=false, p1percent=0, dilate=0, fitluma=false, blur=0, motpn=false, seg=0)
#StackVertical(i, last)
#ConvertToRGB32(matrix="Rec601",interlaced=true)
DeScratch(mindif=1)
ColorYUV(gain_y=10)
SmoothLevels(0, 0.75, 255, 16, 235, Limiter=0, TVrange = false)
Weave()

[poisondeathray]

The despot is not doing anything at all

Your recording fps and the 8mm actual fps isn't the same. There is a mismatch so you have duplicates (recording FPS > 8mm fps)

You have to get rid of the duplicates in order for any temporal spot/dirt removal filters to work properly. They work by looking at differences between frames. If you have duplicates, there are no differences, so dirt/spots aren't "detected" so nothing works

You might use tdecimate in one of the other modes like mode=7 fps=?. I'm not sure what 8mm fps should be (18 ?) . You probably have to bob deinterlace first . If you recorded progressive, it would be better

[mammo1789]

Thanks pois i tried that I deinterlaced using TomsMoComp and QTGMC, I also captured using my canon photo camera 640x480 30 and 25 fps progressive in Mjpeg format recorded on memory card but it compress alot and it losses quality, and get the same result.

Your recording fps and the 8mm actual fps isn't the same. There is a mismatch so you have duplicates (recording FPS > 8mm fps)

You have to get rid of the duplicates in order for any temporal spot/dirt removal filters to work properly. They work by looking at differences between frames. If you have duplicates, there are no differences, so dirt/spots aren't "detected" so nothing works

I thnk that video fred script is doing just that ( assuming that 8mm video has 16 and 18 fps )

You might use tdecimate in one of the other modes like mode=7 fps=?

i didnt see any download site on the author site am I missing something

You mean by doubling the fps with bob how is that gonna help, if there already duplicates

It did clean some big ones but it still something there

I tried something different and video fred script (although says to use progressive video ) I used the same video just flip it vertical ( because the source was up side down) if your theory is true than it shouldn't work here right? but here it is the result ( the best of all concerning anomalies but it blurs and losses detail too much)

[poisondeathray]

You can use other filters to get rid of sthe spots e.g. removedirtmc

I think this is the same frame as the one in your screenshot

The stronger the settings, the more details removed. e.g sometimes things like eyes , or important detalis are removed . But the weaker the settings, the more chance of missing spots. You have to adjust the strength to your tastes like all filters

[poisondeathray]

You mean by doubling the fps with bob how is that gonna help, if there already duplicates

Because there are some fields (not frames) that are unique single fields . If you just throw out blindly, there is less chance of getting proper match (you might inadvertently throw out a good field).

Also the flicker seems to vary on duplicates (if it was a true duplicate field, then you would expect the luminance to be the same). If you bob - you start with everything (all the information) . You have better chance to match the luminance with deflicker algorithms (it can choose from dark or lighter duplicate)

[mammo1789]

Because there are some fields (not frames) that are unique single fields . If you just throw out blindly, there is less chance of getting proper match (you might inadvertently throw out a good field).

Also the flicker seems to vary on duplicates (if it was a true duplicate field, then you would expect the luminance to be the same). If you bob - you start with everything (all the information) . You have better chance to match the luminance with deflicker algorithms (it can choose from dark or lighter duplicate)

I get you know thanks, but as i sad the last screen is from fred script ( he uses removedirtmc) and i left it interlaced TFF didnt do any deinterlacing at all. I think in his script that he uses a lot of blurs and sharpening steps ( unnecessary i think especially for my type of video).

Your screen looks very good can you tell me what did you use, your script ( the parameters that you used and plugins )perhaps

O and i never seems to make to work the NLmeansCL thing although i have the card ( ati 5870 ) to match it

Thanks pois as always very helpful

[poisondeathray]

I just used bob, tdecimate and removedirtmc (the non gpu version) - this was just one way to take care of the spots (there are probably better ways, and you can use a better deinterlacer than bob if you want). You still have to do the other things like flicker

assumebff
bob
tdecimate(mode=7, rate=18)
removedirtmc(50,false)

[2Bdecided]

If you've seen videofred's work then you know what it looks like when 8mm film is transferred properly.

If you care about this film, get someone to do the job properly!

If you're just playing around for fun, by all means have fun - but please understand that the final result is awful, and the best that an AVIsynth script can do is make it slightly less awful.

Cheers,
David.

[mammo1789]

Thanks David

but if you see this link http://www.youtube.com/watch?v=Ea695ldbU_8&feature=related the guy is using camera and not scanned and look at the result and the crispiness of the video it is even way better then fred's default scanned film. Not to mention to process even further in avisynth.

If you care about this film, get someone to do the job properly!

If you're just playing around for fun, by all means have fun - but please understand that the final result is awful, and the best that an AVIsynth script can do is make it slightly less awful.

There is not any that can do that for me ( maybe if I send it to western Europe and pay more for the transport then for the movie).

It seems that my problem is the camera i have cheap 8mm camera and photo camera, the guy is probably using hi def camera and that's why his result looks so amazing.

[filmboss80]

...the guy is using camera and not scanned and look at the result and the crispiness of the video it is even way better then fred's default scanned film.

That is an example of aerial imaging off the projector's film gate, using a DIY "film chain" housing that contains a plano-convex lens and a front-silvered mirror. This is a far cry from using a camcorder to shoot an image being projected on a wall. If you have a good camcorder (CCD imaging chip is far better than CMOS due to rolling shutter issues), you can get some very sharp images. Lenses, mirrors, and diffusion glass for a telecine configuration can be picked up from vendors like Anchor Optics or Edmund Scientific. Diagrams for that DIY housing are usually floating somewhere around the internet.

A variable-speed projector will enable you to get the flicker out. Frosted diffusion glass must be placed in front of the projector bulb to get even lighting across the film gate.

This is a very good way to do real-time, or near real-time, transfers without the expense of sending your films to a high-dollar transfer house. Of course, there are pros and cons to every method, and no film-to-video transfer method is terribly easy.

[2Bdecided]

(CCD imaging chip is far better than CMOS due to rolling shutter issues)

kind of irrelevant for film-to-video transfer - it only matters when something is moving or changing during exposure, and the film frame won't move while you're shooing it! (or if it does, your view of it will be blocked by the projector).

I'm impressed by the aerial imaging clip on YouTube. With any transfer that copies one source frame to multiple video frames, you'll need to use some AVIsynth magic to remove the duplicate frames before you can use other AVIsynth magic like despot usefully.

Cheers,
David.

[2Bdecided]

It seems that my problem is the camera i have cheap 8mm camera and photo camera, the guy is probably using hi def camera and that's why his result looks so amazing.

I think that's an SD camera. It's 4x3, and it's uploaded to YouTube as SD.

[filmboss80]

(CCD imaging chip is far better than CMOS due to rolling shutter issues)

kind of irrelevant for film-to-video transfer - it only matters when something is moving or changing during exposure, and the film frame won't move while you're shooing it! (or if it does, your view of it will be blocked by the projector).

The camera and projector are stationary, but the film is rapidly moving*, and the image on the film itself is constantly changing during the exposure.

http://arstechnica.com/civis/viewtopic.php?f=6&p=22404947

http://www.cinematography.com/index.php?showtopic=36613&st=60

*EDIT: Refers to real-time or near real-time transfers, as mentioned in my earlier post. If one is only transferring at the snail's pace of 3fps, then I agree that the imaging sensor is irrelevant.

[mammo1789]

I think that's an SD camera. It's 4x3, and it's uploaded to YouTube as SD.

It looks like Canon x1 tough. It is interesting to see that when i shot from the set as i posted the camera in front of the projector and very close to beam the picture is relatively sharp, but it has like burn out shadows and contrast ( which is not there on the original) and pixels?! ( it's analog by the way )look for your self I marked the "burnouts" and some sort of alaising pixelisation ( the picture in the white paper is very crisp and contrast balanced( And also it seems that the correcting the levels in proc amp in vdub doesn't solve the "burnout" only makes the whole picture dim or light but not the burned highlights, especially evident on the color footage. This samples are already processed with avisynth and deinterlaced ( so to filters can work)
and when i shot with photo camera and from a distance behind the projector and in to the wall then there is no such anomalies but the picture is slightly blurrier but overall seems better. I borrowed this morning samsung h-mx1 HD progressive 720p camera from my cousin and i will repeat the process to see how it goes. I will also try to build dyi box.
One question tough, after the 45 degrees mirror the beam is hitting what ? plastic, glass, some diffuse material ( i'm talking about the flat surface that you point the camera at the other end of the box

[2Bdecided]

For whatever reason, your camera is utterly crapping out. Sorry, I don't know what the technical term is! One of my old camcorders used to do it under spot lights.

Also, none of your shots are even vaguely in focus.

[2Bdecided]

The camera and projector are stationary, but the film is rapidly moving*, and the image on the film itself is constantly changing during the exposure.

I'll grant you that the guy posting in one of those threads is a genius in this field.

http://www.cine2digits.co.uk/

but...

http://arstechnica.com/civis/viewtopic.php?f=6&p=22404947

That shows two separate original frames of film ending up on one video frame. With CMOS, you'd get all of the unwanted frame, with rolling shutter you get part of it. Neither is in any way acceptable IMO. YMMV!

An acceptable transfer (e.g. 16fps film to 50fps video) has every frame of film existing for the entirely of the exposure of at least one frame/field of video. The film does not move during the video frame. It does not change. Rolling shutter or not is irrelevant in that respect. And you need to find some way of recovering exactly one of each of the original film frames from the video, and dumping all duplicates and blended/merged frames.

Just my opinion. Some people seem happy with dupes and blends. It looks OK on a TV, but is poor (looks ghosty) on a PC, and makes AVIsynth despot and degrain processing hopeless.

Where the rolling shutter does matter is if the blade of the projector goes through the image while capturing. That can cause dark/light banding on the image. That's not mentioned in those threads as far as I can see, but is a real problem (just like flicker - but worse - I agree). You can remove the blades entirely, and delete the resulting blurred pulldown frame in AVIsynth. People have done that, but not me, so I can't comment - it looks like it works: http://forum.doom9.org/showthread.php?t=106837 but maybe not perfectly: http://forum.doom9.org/showthread.php?t=162784

Cheers,
David.

[jagabo]

[QUOTE=filmboss80;2169147]

(CCD imaging chip is far better than CMOS due to rolling shutter issues)

kind of irrelevant for film-to-video transfer - it only matters when something is moving or changing during exposure, and the film frame won't move while you're shooing it!

The film is not moving during the exposure if the projector is working properly. The projector moves the film in jerks. When the blade is allowing light through the film the film is standing still. When the blade is blocking the light the projector moves the film to the next frame. The film is not moving continuously.

[mammo1789]

For whatever reason, your camera is utterly crapping out. Sorry, I don't know what the technical term is! One of my old camcorders used to do it under spot lights.

Nice sad. Yes it seems that is doing some "processing". More intriguing is that today i tried the new hd camera that i talked about, and to my surprise it can't focus and following his advice i used shutter speed 1/50 ( i tried all speeds and on higher speed it catches the screen without the over burn white screen but it rolls black tick lines every second )but the screen looks like over burn all white it cannot focus at all .

Also, none of your shots are even vaguely in focus

. It seems that is the biggest task for my cameras if i put to auto or manual it focuses fine on low moving scenes and stationary, and as soon as high movement scene is on the focus losses it.
On auto tries to refocus while on manual it stays out of focus ( till i do it manual again for every scene which is not practical).
The camera is firmly put the projector is also firmly put it is not that I'm trying to shoot while holding the camera in my hands.

I'm very disappointed (although I knew what to expect) but i went to one of my neighbors and he tried with the same projector ( he has the same "ruse" Russian projector) and with ( watch this ) Benq low quality photo camera 6 mp that can capture video in 640x480 mov and the shots look way sharper and way more focused than mine and most importantly without the burning effect of the Samsung camera that i have, and he puts the camera on tripod and projector on the table projection on white sheet much further than mine setup ( and i think it is better to be closer because mine projected picture is much sharper than his but soon as it is captured by the camera it is different ball game

[jagabo]

If you set manual focus it doesn't make sense that the video is focused when nothing is moving but blurry when things are moving. Manual focus should fix the focusing distance on the screen. It doesn't matter if the picture is changing or not. How far away is the camera from the screen? Are you sure it's just not too close for the camera to focus? Does the camera have a macro mode?

You also need to disable auto exposure on your camera and set everything manually.

[mammo1789]

If you set manual focus it doesn't make sense that the video is focused when nothing is moving but blurry when things are moving. Manual focus should fix the focusing distance on the screen. It doesn't matter if the picture is changing or not. How far away is the camera from the screen? Are you sure it's just not too close for the camera to focus? Does the camera have a macro mode?

You also need to disable auto exposure on your camera and set everything manually.

Yes i know jagabo it doesn't make sense, i tried with the digital camera ( with the analog is doing it in manual I swear like it is defocused somehow and it is not button but slider that you push so it is mechanical i guess ) and it worked but i had to get back some cm far from the "wall" to be in focus. Yes it has macro mode and i put it anything on manual exposure, focus, wb white balance and shooter speed now it is much better but still not there. The main problem is focus and Im working on it the second was blown highlits and it is far less pronounced with the digital camera.

I will try with mirror as somebody sad t see if it is any better or this weekend when i have the time to try with the box

[poisondeathray]

Borrow your neighbor's cheapo camera. It's can't be worse than you are getting now

[2Bdecided]

he puts the camera on tripod and projector on the table projection on white sheet much further than mine setup ( and i think it is better to be closer because mine projected picture is much sharper

No, further away is better because you can get the projector and camera much closer on-axis, reducing keystoning and differential focus/exposure problems across the frame.

There are other issues which favour closer (brighter image, less problem if room isn't totally dark, less reliance on quality lenses to zoom etc) but these are less important, or can be more easily overcome, than the basic problem of trying to project or film way off-axis.

Cheers,
David.

[mammo1789]

Thanks David

It seems that the samsung hd camera can record only in 16:9 bump ( i didn't find anywhere in manual to record in 4:3) even if you select sd resolution.
I played with the exposure to not blow the highlights and in the same time to not be very dark

Also what frustrates me most is that i focus the camera using written words on paper or coin and it is focused ( it is in the middle on the projected picture ) then when i try to record from the projector the picture is blurred go figure.
The samples are not processed pure captures
better than before ?

[jagabo]

Also what frustrates me most is that i focus the camera using written words on paper or coin and it is focused ( it is in the middle on the projected picture ) then when i try to record from the projector the picture is blurred go figure.

Just out of curiosity, what happens if you leave the paper with the words on it on the wall/screen while the projector is running? Does it stay in focus or does it get blurry too?

[2Bdecided]

Is the projector focussed?!

True manual focus won't change - except sometimes on some cameras if you take the zoom to one extreme or the other.

You can manual focus on the moving image itself.

[mammo1789]

Just out of curiosity, what happens if you leave the paper with the words on it on the wall/screen while the projector is running? Does it stay in focus or does it get blurry too?

I tried and it stays focused but it is barely visible because of the low exposure ( needed to not blow the highlights when projects the picture )

The problem with hd camera is also that it captures 16:9 and the film is 4:3 and i get big black bars left and right 1280x720 50p progressive.

Is the projector focussed?!

True manual focus won't change - except sometimes on some cameras if you take the zoom to one extreme or the other.

You can manual focus on the moving image itself.

Yes the lens is movable and i focused it till the picture is bright and focused, I also did build some telecine box ( don't laugh guys ). But I must say that the projected picture on the plastic (the out of the box and yes i measured 45% and played around with the setup to get the best results ) is less sharp than the picture directly projected into white copier paper or the wall from close distance ( it is the most focused and sharp method that i tried so far the only problem with that setup is that then it is very difficult for the camera to get it like the original on the wall is always worse )

[jagabo]

The problem with hd camera is also that it captures 16:9 and the film is 4:3 and i get big black bars left and right 1280x720 50p progressive.

If that gives you a good picture otherwise I suggest you use it. Since you will be filtering anyway you can just crop away the edges of the frame. You'll still have better resolution (960x720) than a standard definition camera can deliver (720x576). And starting with progressive frames is a big plus.

[mammo1789]

Here is sample 1280x720 50p cropped to 860x640 ( so it doesn't have black bars) and opened in avisynth made from mp4 (h264) to hufyv and then opened in fred's script and processed. I get some red highlights in right corner i tough it was from the red light that signals that you are recording but it was duct taped before by me.
This is one of the worst parts of my collection my intention was that if this becomes watchable then the other good parts with ok colors and sharper will look dvd ish.

[mammo1789]

Hm I must be doing something wrong the stationary picture is ok but since is starting to move is all messed up, i know it is something with the fps but im not sure what. I tried to load the h264 mp4 file ) with fps=50 (default=25) and audio false then use tdecimate(mode=7, rate=18) or 16 ( as suggested by fred but my projector has variable speed and how do i know what speed is he rolling ( 16 or 18 or exactly what )

[jagabo]

There is likely some motion blur in the film itself, but your main problem is multiple exposures. Some frames captured by the camcorder contain content two film frames. You can minimize that by using the faster shutter speeds. What you really want to do is coordinate the camcorder exposure with the projector. But you can't do that with your setup.