I have a project right now where I'm using a planar tracker to stabilize a plane in my footage. I then use a paint node to clean it up, and then use the inverse of the planar tracker to reverse this. I then mask this, and merge it over the original footage (to avoid artifacts over all of the image).
The stabilize transform from the planar tracker outputs an image with a large DoD, such that if I feed it back into the reversed planar transform, the result is the size of my original footage.
However, the paint node seems to clip the image to the frame size. This means I end up with only a small area (one that changes size as the plane moves).
Is there any way to avoid this? I couldn't find any options on the paint node.
BTW, I'm using the example from the Blender Guru's tutorial on camera tracking:
Introduction to Camera Tracking, but instead of blender, I'm doing the tracking and compositing in Fusion, and the graphics in Houdini. Once I got past the first frame is zero or one, it seems to work well:
My current animation. Lots of messiness with the marker removal (hence this question), and I'll have to put some work into meaningful reflections and lighting.
MacOS/Linux, Divinci Resolve 16, Fusion 16.