Regarding the Raindrop detection algorithm

[RamitPahwa]

Amazing Work,!
I have the following understanding when we inverse binarize the image, we will also brighten the darker region of the image which did not have gradients to begin with, how do you just segment out the raindrop in such a scenario?

The example you mention in the paper also has no gradients in the sky region and the bottom right as shown, binarizing the image should segment the sky along with raindrop, how do you seperate them, is there any additional step involved?

Could you please share you script for the raindrop detection if possible, it would greatly accelerate my [research.]

@V-Soboleva / @oleg-Shipitko

[oleg-Shipitko]

@RamitPahwa thank you for feedback!

The algorithm may misinterpret homogeneous and static regions such as the sky. However, as we have seen in our experiments, sky regions usually have small gradient value and can be separated from raindrops by applying Gaussian blur before binarization and by adjusting the binarization parameters.

Unfortunately, I can't share the code, but I hope my answer will help with your research. Good luck, and thanks for your interest in our work!