End-to-end algorithm for absolute phase retrieval

Abstract

Absolute phase plays a very important role for many applications. At present, various techniques have been exhaustively developed for absolute phase recovery. There are three major categories: spatial phase unwrapping, temporal phase unwrapping and deep learning techniques. Analyzing from these methods, a commonality is limited into the procedure of absolute phase recovery. Namely, the conventional algorithms yield to calculate wrapped phase, fringe order and absolute phase in turn. There is a closer dependency between the accuracy of the previous results and the next ones. So more serious errors could accumulate from this computation procedure, meanwhile the recovery speed is decreased. For this reason, we present an end-to-end method to obtain the absolute phase maps by projecting only two-step specially intensity-coded fringe patterns. Unlike the above-mentioned techniques, we can not only compute the wrapped phases from two fringe images, but also decode the corresponding fringe orders simultaneously from two fringe images without any additional patterns. The presented method breaks the limitation and simplifies the procedure of phase unwrapping. Simulations have been carried out to validate the feasibility of the proposed method.