Slit-scan: Is it better than my video converter? Can it better 50 FPS?

[tryit]

Can you tell me all about ML and slit-scan images?

I created my own slit-scan software, but have since heard ML can do this.

Knowing nothing about ML or details of camera internal operation - how does ML do slit-scan - can ML change how info is obtained from the sensor? e.g. can slit (or frame) be read more than 50 FPS?

I use a canon 5D MkIII.

My program converts any .mov file (from any camera) to a slit-scan image with some controls on slit position, scanning etc.  The main limitation is the camera's FPS rate which I would dearly love to be much greater than 50FPS.

I don't really know how normal camera video mode works, but it seems that there is a non interlaced raster scan that takes 1/50 of a second (or 25 or 24), and each pixel must be exposed for some fraction of this time(?) as there is a lot of motion blur (can this be controlled?) on a frame of a moving image and a vertical moving horizontal line is sheared (becomes angled).

See www.slitcam.com where you can download and try the program - let me know what you think.



Thanks!

[kihlbahkt]

Cant really comment on slit scan but you wont be able to get 50fps out of the cam with ML. The FPS override feature in ML is meant for reducing FPS not increasing it.

[a1ex]

What about using something like Interframe or Twixtor to double the frame rate?

ML can do smooth slit-scan pictures, but not video. Set FPS override to 2 or 3 fps, high jello mode, shutter speed as fast as you can, start recording video, then extract the frames you like.

[tryit]

I haven't tried ML yet. I am interested in how ML works - my 5D can produce 50fps video, can ML (even if theoretically) better that either for video or slit-scan (or other mode?) or is 50fps a limitation of the sensor? I note some p+s cameras do boast very high frame rates, but quality and number of pixels is apparently much reduced.

It looks like Interframe or Twixtor just interpolate between existing frames - you cannot add information, this would be the same as (or similar to)  stretching the image my program outputs in photoshop (bicubic interpolation).

I did have an idea that photographing via a set of mirrors parallel with the raster (so the same long thin image is repeated on the sensor)  you could take more than one line of pixels off each video frame. Should be an improvement, but I suspect rather marginal and a lot of effort.

Although I was delighted with the image I posted (it was my first attempt!) clearly it would be improved with a faster framerate. Photofinish cameras obviously have fast "frame" rates, but are presumably rather specialist bits of kit and I hoped to get what I could from my existing equipment.

Does slow slit scan in ML use the full sensor width/resolution or is it just full HD video size. Is slit-scan on the 5D iii release?

[daimttery]

Video Converter - REMOVED