8mm Scanner

Guest · Post by **Guest** » Mon Jan 03, 2005 2:23 pm

I just got working a system for converting Super-8mm/Reg 8mm to DVD frame by frame.

http://www.jiminger.com/s8/index.html

I'm looking for suggestions on improving the process.

Thanks
Jim

Post by **jimfcarroll** » Mon Jan 03, 2005 2:25 pm

My apologies on that last post. I thought I was logged in.

I am now.

Jim

Post by **Petteri** » Mon Jan 03, 2005 3:12 pm

Wow....

May i ask how long it took to scan all images in that sample ;)

Can you also post one fullframe sample?

Petteri

Post by **timdrage** » Mon Jan 03, 2005 3:22 pm

Looks pretty impressive! Bonus points for use of LEGO!

Wouldn't the inter-strip correlation malarky have trouble with moments of no movement in tripod shots though? or is it accurate enough to detect frames with identical grain?

Anyway, nice work!

Post by **jimfcarroll** » Mon Jan 03, 2005 4:01 pm

Petteri ,

The scanner is slow. It takes about 8-10 hours for a 50 foot roll. It takes upwards of 3 days (sometimes less) for a 400 foot reel. I'm guessing that clip is about 1/2 a 50 foot roll (it's actually an excerpt from a 400 foot reel) so that scan time was about 4-5 hours for that much footage. The image processing runs in parallel and is faster than then scan so all of the individual frame images are ready to go as soon as the scan is done.

timdrage,

I don't have any footage with the camera on a tripod filming a completely stationary scene so I suppose it could be a problem in that case. As it is it works fine with all of the footage I have. The biggest problem with the correllation is when overlap is missing completely (it still makes a "best guess").

Currently my BIGGEST problem is that the film will occationally move during the scan causing a break or shift along the entire length of the image. THIS casuses the most trouble for the correllation since the end of one strip is distorted. I'll also have a problem with the correllation if the sprocket is not detected perfectly in two successive strips. In a 400 foot roll this happens 3-4 times so it's not that frequent. I'm working on making the sprocket hole finding more robust to eliminate that.

Thanks,
Jim

Post by **jimfcarroll** » Mon Jan 03, 2005 4:19 pm

Petteri,.

I'll put a few full frame JPEG samples out there tonight.

Jim

Post by **jimfcarroll** » Tue Jan 04, 2005 3:33 am

Per the request, I added a couple of sample frames to my site. I do not link to them from my main page but you can see them at these URLs:

http://www.jiminger.com/s8/Sample1.jpeg
http://www.jiminger.com/s8/Sample2.jpeg

Post by **paulcotto** » Tue Jan 04, 2005 4:39 am

My film perforator uses a servomotor and a gecko servo driver to move the film. You could also use stepper motors to do the same thing with a reduction drive. most steppers go 200 steps per 360 deg. rotation. Drop that down 10:1 and you can move your film very precisely. I use a cheap laptop to run my machine by the parallel port. I use CNC software to handle the motion control. Here are some links>>

http://www.dakeng.com/turbo.html

http://www.geckodrive.com/

https://www.sdp-si.com/eStore/default.asp

http://www.lavezzi.com/1667Pitch.html

http://www.herbach.com/

http://www.meci.com/Catalog/Category/91 ... 1f8d7e15db

Regards,
Paul Cotto

Post by **jimfcarroll** » Tue Jan 04, 2005 1:13 pm

Thanks for the links. I was going to do this originally with a stepper motor but I found them difficult to work with (I'm not very good at either mechanics or electronics). My initial attempt at the film forwarder began by pulling the stepper and controller from an old 5 1/4" floppy drive. Now that I've had some success with a DC motor maybe I could try the stepper approach again. Any suggestions on specific controllers/motors?

Thanks
Jim

Post by **VideoFred** » Tue Jan 04, 2005 2:11 pm

Jim,

I must admit: it looks very good. I didn't expect this quality from a scanner.

But maybe you can make a divX-avi of your original? Then we can see the mpeg in real size...

Edit: I took a better look at your site, I'm very impressed about your knowledge both general and about writing software. How do you deal with the framerate pulldown? I think you can help me with my Avisynth script to do the pulldown. (See my 'auto-processing' topic)

Fred.

Post by **mattias** » Tue Jan 04, 2005 4:16 pm

i don't know how you do the sprocket detection exactly, but finding two holes, one on each side, and then linearly interpolate the others should be faster and result in perfect registration assuming that the film was properly perforated.

one question: the clip has too much contrast while the stills have very little. was this intentional? also the stills aren't sharp. the scanner i'm using shows the grain clearly when i scan a strip of film. it's a "neg scanner" but a flatbed shouldn't be far behind these days? perhaps it's not focused correctly or you're not actually scanning at the high resolution you think?

/matt

Post by **christoph** » Tue Jan 04, 2005 5:27 pm

mattias wrote:it's a "neg scanner" but a flatbed shouldn't be far behind these days?

hmm.. what makes you think that a scanner that costs under 300 dollars and that was designed to scan reflective surfaces in an area like 200x300mm could compete with one that was designed to scan only transparent media from a area of 36x24mm? and costs two or three (or four) times as much?

++ christoph ++

Post by **jimfcarroll** » Tue Jan 04, 2005 5:43 pm

mattias wrote:i don't know how you do the sprocket detection exactly, but finding two holes, one on each side, and then linearly interpolate the others should be faster and result in perfect registration assuming that the film was properly perforated.

I use a linear interpolation using: 1) the location of the "most likely" sprocket hole, 2) the known scanner resolution (the samples are 3200 dpi), and 3) the location of the edge of the film, to help with the selection of other sprocket holes. This allows me to effectively separate peaks in the Hough transform that correspond to actual sprocket holes, from noise in the transform. Once I locate the appropriate peaks and I know which edge pixels correspond to the sprocket holes in the image, I "fit" a model of the sprocket hole to the known edges pixels. This gives me an exact scale, rotation, and center location for each sprocket hole.

This was not enough to pull the exact frame out though. There was too much slop in those parameters to extrapolate the position of the frame. Originally I did it this way but the frame kind of floated around as the movie played. Instead I thought about how a real gate works and now I use the center position of the sprocket and a perpendicular line to the far edge.

mattias wrote:one question: the clip has too much contrast while the stills have very little. was this intentional? also the stills aren't sharp. the scanner i'm using shows the grain clearly when i scan a strip of film. it's a "neg scanner" but a flatbed shouldn't be far behind these days? perhaps it's not focused correctly or you're not actually scanning at the high resolution you think?

The movie clip and those sample frames are not from the same reel. I deleted all of the extracted frames that made up the reel that the movie clip came from. I can reextract some of the frames from that sequence and put them out tonight.

Since the dynamic range of my scanner is pretty crappy (what else would you expect for $200) I don't like to do a lot of post processing to change the contrast. I've been fiddling with different scanner contrast settings (mostly via the histogram manipulation) but I don't have much experience with what would be correct aesthetically and I'm sure the reel that the movie clip came from had scan setting very different from the ones that the sample jpegs came from. Suggestions are certainly welcome.

Jim

Post by **jimfcarroll** » Tue Jan 04, 2005 5:51 pm

One more thing - I thought about a Film Scanner (I wanted to try a Nikon Coolscan) but I couldn't get hold of one for testing and I wasn't going to lay out the $600+ on a hunch. Since it's a closed box I didn't know if I could pull the film throught it. If it's faster and I can actually do it I might consider it. If anyone has any suggestions, please let me know.

Jim

Post by **mattias** » Tue Jan 04, 2005 6:03 pm

christoph wrote:hmm.. what makes you think that a scanner that costs under 300 dollars and that was designed to scan reflective surfaces in an area like 200x300mm could compete with one that was designed to scan only transparent media from a area of 36x24mm? and costs two or three (or four) times as much?

oh, i didn't think it would produce as good images, but if it says 3000 dpi you should get that, shouldn't you?

/matt