statistical outlier removal isn't enough

[soumya997]

Hi guys, I noticed there's a new denoising feature available.

Previously, I used statistical outlier removal from Open3D on the extracted ply file, which wasn't much effective, struggled with dense floaters generated in 3DGS. These floaters are usually very dense, which I believe makes it difficult for general algorithms to remove them effectively.

I was wondering if there might be a better approach to tackle these floaters. If they can be successfully removed, converting the point cloud file back to a 3DGS file should also eliminate the floaters there. Below are some examples where I'm encountering this issue.

img 1	img 2