About calculating mask #4
Description
Hey,
I am curious about the real implement of the calculation of the mask.
In the supplemental, the mask is the product of two binary masks:
The first is based on the previous paper, and the other is based on the warping residue.
However, i find three mask in the code: exclusive_mask1*Mask1*warped1.
So i want to know what is the difference between the mask calculated in collision_mask folder and the original method in model-blind video denoising via frame-to-frame training.