|
Post by dennywilson on Mar 7, 2012 22:22:51 GMT
So you might as well do what I suggested: replace the region of the frame in which the double-imaging occurs with the off-air video. It's much simpler, doesn't require precise alignment, doesn't require precise matching of gamma, and gives results that are just as good (quite possibly better, because the noise originates only from the off-air image, not from both as it would using your technique). Richard. But is this possible? Wouldn't it be easer to remove the double frame and recreate the missing frame by taking the good frames on either side and run it through motion estimation and create a "new" frame with elements of the frames on either side?
|
|
|
Post by Richard Russell on Mar 7, 2012 22:40:50 GMT
If the background is static, you could use a neighbouring frame from the film recording to replace the double-imaged area. That way you avoid degrading the picture quality. You could perhaps do that, but only on the first and last frame of the 'moving' sequence. In the intermediate frames there is nowhere from which you can copy the 'clean' video you need to replace the double-imaged area (since the adjacent FR frames will themselves contain a mixture of two original frames). From the point of view of alignment you can do either, but it's obviously preferable to end up with an 'undistorted' final image rather than a 'distorted' final image. So since the off-air (VT) image has suffered no distortion, whereas the FR is distorted, it's preferable to align them by removing the distortion from the FR rather than by adding distortion to the video. In case you don't think it makes much difference, look at the Colour Recovered Dad's Army episode. Whenever there's a pan you can clearly see the background stretching and shrinking as it crosses the field of view. It's quite disconcerting! Richard.
|
|
|
Post by dennywilson on Mar 8, 2012 1:56:23 GMT
So there is no hope of dealing with the double-frame issue then?
|
|
|
Post by Alex Weidmann on Mar 8, 2012 17:02:02 GMT
If the background is static, you could use a neighbouring frame from the film recording to replace the double-imaged area. That way you avoid degrading the picture quality. You could perhaps do that, but only on the first and last frame of the 'moving' sequence. In the intermediate frames there is nowhere from which you can copy the 'clean' video you need to replace the double-imaged area (since the adjacent FR frames will themselves contain a mixture of two original frames). I'm not sure how SVS do it? Possibly they paint through to a frame where the static background is fully revealed (once the foreground object has moved out of the way); rather than to an adjacent frame. That could work in some instances, provided the camera-shot is locked-off. Yes I noticed that. Also saw the same effect on "The Space Museum". I wonder whether you think the stretch in the right and left margin is consistent enough to apply the same (or similar) undistort transform across different FRs? Would be nice to have better geometry on the Hartnell and Troughton episodes. Only thinking about the gross distortion ofcourse, not the more complex fine-scale distortions measured by your CR process. By the way, I was watching the "Terror Of The Autons" DVD yesterday, and noticed that episode 4 seemed to have obvious line-structure on the Autons bright yellow jackets. I checked the FFT, and saw quite a strong signal at 0 Hz, 288 c/aph. The spectral qualities are quite similar to the TOTP clip: so I ought to be able to get field-separation on the out-of-phase film inserts in that episode. (The other episodes have very little energy at 288 c/aph unfortunately.)
|
|