This means that a constant-t slices of an object's worldline will simply differ by translations, while constant-x/y slices differ arbitrarily if the object's speed changes.
Hence, motion compensation for 1D time + 2D transforms for 2D space is the optimal approach.
Fundamentally, this is because almost all videos depict evolving physical reality (having a Lorentzian metric) with 3D space projected to a 2D plane and with objects moving at non-relativistic speed relative to the camera, which means that:
1. Due to Lorentzian metric, each object's own time is one dimensional
2. Due to non-relativistic relative speeds, the time evolution vectors of them all are approximately always the same as the camera, and thus there is approximately a single "time" vector for all objects, meaning that you won't observe any relativistic length contraction of objects
3. Due to camera projection of 3D space, there are two other space dimensions
Copyright © 2017, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds