Abstract

Universal photometric stereo (PS) aims to recover high-quality surfacenormals from objects under arbitrary lighting conditions without relying onspecific illumination models. Despite recent advances such as SDM-UniPS and UniMS-PS, two fundamental challenges persist: 1) the deep coupling between varyingillumination and surface normal features, where ambiguity in observed intensitymakes it difficult to determine whether brightness variations stem fromlighting changes or surface orientation; and 2) the preservation ofhigh-frequency geometric details in complex surfaces, where intricategeometries create self-shadowing, inter-reflections, and subtle normalvariations that conventional feature processing operations struggle to captureaccurately.