When using the C_ColourMatcher node in CaraVR, you may notice that the result after analysis does not appear better than the initial solve, and you may also see a banding effect created.
As example, here is a problematic colour matching result:
C_ColourMatcher works by matching the whole of the camera input and if you have a round lens with a black border, then this can offset the colour matcher making the results worse.
To avoid the above issue you should:
- create an alpha for your camera inputs by rotoing around the image to avoid including any black areas
- point the C_CameraSolve node to use the alpha created above for the Mask shape parameter
The C_ColourMatcher node will start using the same alpha when analysing the exposure and colours and will ignore the black part of the image during processing. This should help improve the C_ColourMatcher node's accuracy.
Additionally, CaraVR 1.0v2 introduced Multi-Band blending mode in the C_Stitcher node. Setting the Blend type to Multi-Band can also help reduce banding.
As example, here is the final colour matching result with fixes applied:
NOTE: All images are courtesy of Universal Post