Learned Nonlinear Predictor for Critically Sampled 3D Point Cloud Attribute Compression

Abstract

Translate Article

We study 3D point cloud attribute compression via a volumetric approach: assuming point cloud geometry is known at both encoder and decoder, parameters θ of a continuous attribute function f : R 3 → R are quantized to θ and encoded, so that discrete samples f θ (x i ) can be recovered at known 3D points x i ∈ R 3 at the decoder. Specifically, we consider a nested sequences of function subspaces is a family of functions spanned by B-spline basis functions of order p, f * l is the projection of f on F (p) l