Synnove
4th December 2017, 22:12
I have two HD spec clips that are edit identical. They are different scans/masters of the same film, so the content is rather similar. The differences are the color grading and the frame geometry.
The A transfer is much much more detailed, crisp, and free of compression artifacts than B. However, the shadow detail has been crushed out of existence by a revisionist color-grader. I need to align the frames so I can combine the best attributes of both (I have the combination procedure finished, I just need to work out getting them aligned).
The geometry differences are nonlinear so I can't use something like Nuke's F_Align or Photoshop's align features, and they aren't constant through a sequence due to gate weave and the like. The framing is also slightly different.
I've an interleaved clip with the masking bars removed ready to go.
https://forum.doom9.org/showthread.php?t=154739 <- This individual seems to've had success but I've not been able to replicate it. It seems like the MVTools solution would work best but, when I tried to replicate pbristow's method, it turned in to a jello-esque dessert. um3k's solution aligned them okay, but it couldn't account for the nonlinear differences so they were at best aligned in the center only.
The A transfer is much much more detailed, crisp, and free of compression artifacts than B. However, the shadow detail has been crushed out of existence by a revisionist color-grader. I need to align the frames so I can combine the best attributes of both (I have the combination procedure finished, I just need to work out getting them aligned).
The geometry differences are nonlinear so I can't use something like Nuke's F_Align or Photoshop's align features, and they aren't constant through a sequence due to gate weave and the like. The framing is also slightly different.
I've an interleaved clip with the masking bars removed ready to go.
https://forum.doom9.org/showthread.php?t=154739 <- This individual seems to've had success but I've not been able to replicate it. It seems like the MVTools solution would work best but, when I tried to replicate pbristow's method, it turned in to a jello-esque dessert. um3k's solution aligned them okay, but it couldn't account for the nonlinear differences so they were at best aligned in the center only.