Study on the improved CGAN based method for the separation of aliased fringes

Traditional Phase-Shifting Profilometry (PSP) requires at least three fringe patterns to project, with accuracy improving as the number of fringe patterns increases. However, using too many fringe patterns can prolong the projection and acquisition time. This paper proposes a deep learning-based method for separating aliased fringe patterns. First, it reduces the number of fringe patterns through entanglement, and then utilizes deep neural network algorithms to separate the entangled fringe patterns, resulting in fringe patterns that are nearly identical to the projected fringe patterns. Experimental results demonstrate that using just two entangled fringe patterns can achieve high-quality 3D reconstruction.