graphdeco-inria / diff-gaussian-rasterization

Other
834 stars 257 forks source link

About the low-pass filter #14

Open lzhnb opened 11 months ago

lzhnb commented 11 months ago

https://github.com/graphdeco-inria/diff-gaussian-rasterization/blob/59f5f77e3ddbac3ed9db93ec2cfe99ed6c5d121d/cuda_rasterizer/forward.cu#L108-L112

According to the equation 33 in EWA Splatting, this code is used to conduct low-pass filter by modifying the cov mat. And EWA Splatting said the low-pass filter is represented by a Gaussian with identity cov mat. So why add 0.3 here? And why 0.3 is related to a pixel?

Thanks!

InFistLee commented 8 months ago

I have the same question here about why scaling the Identity matrix with 0.3. I assume 0.3 is for the 3 sigma principle but not sure. Do you have any idea?

nnop commented 1 month ago

To make a gaussian 1 pixel width, the 1sigma could be half pixel. So the covariance was added by 0.3 (≈0.548^2).