|
Post by Alex Weidmann on Feb 28, 2012 10:41:28 GMT
Does anybody know how to convert a video file into this .y format? (Ideally in a software such as After Effects) AviSynth is what you need, along with Virtual Dub. Split your video file into a targa sequence using Virtual Dub, then read the following thread: forum.doom9.org/archive/index.php/t-149167.htmlIt's not particularly straightforward, but it works.
|
|
|
Post by Alex Weidmann on Feb 28, 2012 10:53:02 GMT
I'm sure it wouldn't be difficult to get you some sample frames of out-of-phase film inserts, there are plenty of them! That would be good, but undistorted samples would be best, as my own undistort process is painfully slow and laborious!
|
|
|
Post by Richard Russell on Feb 28, 2012 11:10:33 GMT
That would be good, but undistorted samples would be best, as my own undistort process is painfully slow and laborious! Well, if it's from a Film Recording it will be distorted! I could send you something that's been through my 'undistort' process, but that's nowhere near accurate enough to distinguish odd lines from even lines by their position. I thought the whole point of your process was the ultra-accurate undistortion operation, after which it's a simple matter of 'separating' the fields with a vertical filter. Have I misunderstood? Richard.
|
|
|
Post by Alex Weidmann on Feb 28, 2012 16:18:02 GMT
That would be good, but undistorted samples would be best, as my own undistort process is painfully slow and laborious! Well, if it's from a Film Recording it will be distorted! I could send you something that's been through my 'undistort' process, but that's nowhere near accurate enough to distinguish odd lines from even lines by their position. I thought the whole point of your process was the ultra-accurate undistortion operation, after which it's a simple matter of 'separating' the fields with a vertical filter. Have I misunderstood? Richard. You haven't misunderstood; I was just hoping your undistort process was as accurate as mine, because it's about 21 times faster from what you told me! In a correction to what I wrote earlier: the TOTP clip has a very strong signal at (0 Hz, 288 c/aph): the signal strength is about the same as the chrominance frequencies. The other source I tested (containing the out-of-phase film insert) has very little energy at (0 Hz, 288 c/aph), hence why it failed to separate. It also has less bandwidth overall, and a much more attenuated signal at the high frequency end. So a 2D Fourier Transform will be a very good indicator of whether my process will work or not. I'm going to have a look at some 377i DVD's next to see if I can find much energy at (0 Hz, 188.5 c/aph).
|
|
|
Post by Richard Russell on Feb 28, 2012 16:57:55 GMT
You haven't misunderstood; I was just hoping your undistort process was as accurate as mine, because it's about 21 times faster from what you told me! No, it's nowhere near as good, not least because it only works where there is really good chroma from which I can extract reliable X and Y displacement values. Also, as I think you know, it's fitting the X and Y distortions to polynomials of the following form: A+Bx+Cy+Dx 2+Exy+Fy 2+Gx 3+Hx 2y+Ixy 2+Jy 3+Kx 4+Lx 3y+Mx 2y 2+Nxy 3+Oy 4Of course for the purposes to which my undistort process is put, it doesn't need to be very accurate. If your experiments would suggest a different polynomial would be a better choice (hopefully with no more terms) I could easily try it. I knew that must be the case. Unfortunately it's not typical. Richard.
|
|
|
Post by rdenham on Feb 29, 2012 21:03:23 GMT
Does anybody know how to convert a video file into this .y format? (Ideally in a software such as After Effects) AviSynth is what you need, along with Virtual Dub. Split your video file into a targa sequence using Virtual Dub, then read the following thread: forum.doom9.org/archive/index.php/t-149167.htmlIt's not particularly straightforward, but it works. Thanks, worked a treat!
|
|
|
Post by Brad Phipps on Feb 29, 2012 22:02:35 GMT
A+Bx+Cy+Dx 2+Exy+Fy 2+Gx 3+Hx 2y+Ixy 2+Jy 3+Kx 4+Lx 3y+Mx 2y 2+Nxy 3+Oy 4Oh... that's where I went wrong, I forgot to carry the K.
|
|
|
Post by Alex Weidmann on Mar 1, 2012 7:57:04 GMT
If your experiments would suggest a different polynomial would be a better choice (hopefully with no more terms) I could easily try it. My method uses a grid, with a different fitting-polynomial for each grid square. They're all cubic functions, with various different options for the number of grid squares. 64X64 is the most accurate. What do SVS use to register their sources? Is it some kind of DVE? It must be fairly accurate to conform film recordings to the off-air Umatic and Betamax sources. I knew that must be the case. Unfortunately it's not typical. I bet you'll find that all the FRs with out-of-phase inserts have no energy at 288 cycles! I was wondering whether out-of-phase inserts have been repaired in the past using reverse standards converted 525 quad tapes? Since this process recovers the original PAL fields, it ought to be possible to fix the field-order of the inserts shouldn't it?
|
|
|
Post by Richard Bignell on Mar 1, 2012 9:25:17 GMT
I have to confess that I find the whole process magical and mind-boggling, especially with the mathematics concerned, but it's brilliant that people are continually trying to find ways to make the process work better. I have to admit that I feel very much like Charlie Brown when talking to Lucy and Linus... ;D
|
|
John Wall
Member
Posts: 4,145
Member is Online
|
Post by John Wall on Mar 1, 2012 13:26:30 GMT
Clarke's third law:
"Any sufficiently advanced technology is indistinguishable from magic."
|
|
|
Post by Richard Russell on Mar 1, 2012 14:16:24 GMT
My method uses a grid, with a different fitting-polynomial for each grid square. I need to use 'global' polynomials so that, even though I have no information about many of the blocks, I can estimate how much to move them without introducing any 'creases' in the undistortion field. Smoothness of the undistortion is more important than accuracy (both to the CR process and subjectively). As I understand it they never do that. Because the 'DVE' isn't particularly high quality they warp the chroma (which is non critical) to match the luma (of which you want to maintain the very best quality). Of course this means that although they are fairly well registered both are distorted, which you can sometimes notice during pans and the like. With the more recent work like Ambassadors it's been done quite differently. I've 'undistorted' the luminance as a side-effect of the CR processing and that has been overlaid onto the chrominance. Little if any further correction has been required to achieve acceptable registration. I've no idea. I would guess the acceptability of this would depend on whether the sudden shift in picture quality from the Film Recording to/from the RSC would be even more offputting than the superimposed fields. Richard.
|
|
|
Post by Alex Weidmann on Mar 2, 2012 13:38:02 GMT
I've no idea. I would guess the acceptability of this would depend on whether the sudden shift in picture quality from the Film Recording to/from the RSC would be even more offputting than the superimposed fields. I think 525 quad tapes are used in preference to film recordings: so where such a quad tape exists, the whole episode will be a reverse standards conversion. I imagine any out-of-phase film inserts would have been fixed if the recovery of the original PAL fields was good enough to allow for it. What I'm not clear on, is what has been considered where the only surviving 525-line copy is sub-broadcast quality. If the Reverse Standards Converter works on Betamax and Umatic tapes, I imagine you could repair any out-of-phase film inserts using these sources; but the results may have been considered too soft to be acceptable. Or perhaps too jarring when transitioning from the film recording. Though I would've thought you might just get away with it from a Umatic source, as these had a fairly decent luma bandwidth. On the other hand, it may be that recovering the PAL fields from off-air NTSC Umatic and Betamax recordings is not viable at all, due to the low-pass filtering of the off-air signal when it was recorded.
|
|
|
Post by Simon Broad on Mar 2, 2012 15:54:50 GMT
I think this thread should be closed as the thread is going on and on because that is what the last one did.
|
|
|
Post by Richard Bignell on Mar 2, 2012 16:11:25 GMT
It's nothing of the sort, Simon.
|
|
|
Post by Rob Moss on Mar 2, 2012 16:43:54 GMT
I think this thread should be closed as the thread is going on and on because that is what the last one did. Who made you a moderator..? This thread is going on and on, but ON TOPIC, unlike the last one..!
|
|