Hello author~ I'd like to ask a question as when computing the homograph warp to compute the multiview consistency, why use alpha-blended depth and normal instead of using the intersected depth & normal with median splats? Just as the GeoNeuS mentioned, only focusing on the splat with the max weight along the pixel ray can reduce the influence caused by the bias in alpha-blended rendering, and lead to a smoother geometry.
Hopefully, this advice could help to get a better quality geometry~
Hello author~ I'd like to ask a question as when computing the homograph warp to compute the multiview consistency, why use alpha-blended depth and normal instead of using the intersected depth & normal with median splats? Just as the GeoNeuS mentioned, only focusing on the splat with the max weight along the pixel ray can reduce the influence caused by the bias in alpha-blended rendering, and lead to a smoother geometry. Hopefully, this advice could help to get a better quality geometry~