This chapter looks at the removal of unwanted objects (rigs) from image sequences without accurate rotoscoping or keying to produce a clean plate.
In this context, we define a rig as a foreground element in a sequence that moves over a background element. The plug-in will only work satisfactorily if it is possible to model the background motion by a global 3D transform. For example, if the background contains multiple objects moving in different directions, the results will be poor. Typically, good results will only be achieved in situations where a skilled artist could generate, and track in, a clean plate in order to repair the sequence. However, this plug-in should make the process quicker and easier. The rig removal algorithm works by estimating the background motion between successive frames, ignoring the foreground object, and then using the motion information to look forward and backward in the sequence in order to find the correct piece of background to fill in the missing region. The size of the missing region and the speed of the background dictate how far away from the current frame it is necessary to search to find the correct information.
|
|
|
| Before with bird. | After applying F_RigRemoval. |
|
|
|
| Before with taxi. | After applying F_RigRemoval. |