from 25 progressive to 50 interlacedsuper8 footage
Posted 19 December 2012 - 06:37 AM
the goal is to lower the grain appearance of super8 telecined footage at 25 fps . if i put the videotrack on the timeline, then overimpose the same video track but shifted of one frame and with opacity set at 50% i got a smoother grain due to averaging of frames ( In fact i am mixing the same video track with itself). If the camera and object is stationary this system works fine,the grain is very close to 16mm, moreover there is a sort of stabilization. but if there is panning or movement i get a sort of ghosting/strobo effect of course. NOW, is it possible to weave , or for better to say, interlace these frames ? In this case i shouldn't get ghosting effect , moreover moore smooth in movement action. Another improvement should be higher vertical definition. In few words, we pass from 25 progressive to 50 interlaced, isn't it ? I am utilizing powermac g5 and FC studio2 (FCP 6)
Posted 19 December 2012 - 08:13 AM
Seems to me first thought as that you're always going to have some kind of motion weirdness with a post interlace effect (there is even weirdness with interlacing at acquisition ...
Maybe have a think about shooting at 50fps progressive then blend to 25fps progressive.
Funny thing in that case is that you'd end up using twice as much film per frame, and probably get the same result in grain reduction as if you'd used twice as much film area in the spatial sense (i.e. a larger frame as opposed to two sequential temporal frames).
I wonder if PJ's hobbit could achieve lower noise in the blacks by this approach in any 24fps presentations ?
Posted 19 December 2012 - 09:51 AM
Posted 19 December 2012 - 10:00 AM
Posted 19 December 2012 - 10:53 AM
Posted 19 December 2012 - 12:10 PM
Posted 19 December 2012 - 03:13 PM
Hmm, are you sure you weren't mistakenly applying a filter in an interlaced environment? They can end up sort of interpolating between frames, sometimes, which does wreck a progressive look.
Isn't that what Roberto outlined in the first post?
Roberto, while I'll agree sometimes there are 'something for (next to) nothing' algorithms out there (hashtables for instance) - thing is there is usually an element of amortised cost involved (hashtables) or the gain you've achieved has been made at the cost of another factor which is usually hidden/or they fall below a threshold in our senses (compression etc...).
Your idea as far as I can understand it so far is trying to do is break the rule of:
more film = less grain
(all the while keeping other things equal, such as the 'film look')
Yes, it'll work, two ways - use more film in the spatial sense, 16mm vs. 35mm - use more film in the temporal sense, 24fps vs 48fps with frame blending
As for another method that uses the same amount of film in the first instance - good luck!
If you do manage to crack it careful you don't ruin your patent rights by publishing the results here
Aaton digital has an interesting and as far I know new approach, though its obviously off topic