CaliSplat: Efficient Gaussian Image Compression via Rate Allocation
November 4, 2026
CaliSplat implements a sparse Gaussian fitting method for image compression using a calibrated rate allocation strategy. It uses closed-form reconstruction error to identify high-complexity regions, fitting more Gaussians only where necessary to optimize a constrained capacity budget.
HOW THIS AFFECTS YOU
●
builderYou can potentially achieve faster, deterministic decoding for image assets without heavy neural networks.
●
researcherThis offers a new approach to optimizing 2D Gaussian representations through local difficulty estimation.