Machine learning methods for digital holography and diffractive optics
Pavel A. Cheremkhin, Nikolay N. Evtikhiev, V. V. Krasnov, Vladislav G. Rodin, Dmitry A. Rymov, Rostislav S. Starikov
Abstract
With active advancements in computer and computational technologies, deep learning has found its way into many fields. Recently it has become an active topic of research in diffractive optics and holography. Deep leaning techniques have been shown to benefit greatly from abundant information offered by using both amplitude and phase of the optical field. These techniques can be applied for image reconstruction, zero-order suppression, hologram generation, etc. In this paper various learning based methods for enhancing digital and computer-generated holography are analysed. We demonstrate a deep learning model for generating diffractive optical elements from an arbitrary intensity-only image. Numerical evaluation of model’s performance has shown that generated diffractive optical elements are of acceptable quality.