A weighted least squares approach for phase retrieval based on patch-wise noise level estimation

Phase measurement profilometry (PMP) is used in a wide range of 3D optical metrology applications. It is highly resistant to ambient light and object albedo, but it can be more prone to sensor and environmental noise. In this study, we present a novel approach for retrieving phase from noisy fringe patterns. Our method is based on weighted least squares (WLS) and noise estimation techniques. Firstly, we divide the noisy fringe patterns into overlapping patches. Then, we apply principal component analysis (PCA) to the patches, which allows us to effectively estimate the noise level and weight. We analyze the weight estimation and phase retrieval capability of our method concerning modulation, shift step, frequency, and size of the fringe pattern. We also provide simulation and experimental results to demonstrate the practical applicability of our proposed method.