Quantitative phase imaging via a cGAN network with dual intensity images captured under centrosymmetric illumination

Abstract

We report an end-to-end approach for quantitative phase imaging based on two intensity measurements. In our approach, we sequentially illuminate the thin complex object using two centrosymmetric LEDs whose incident angles are close to the maximum acceptance angle of the objective lens. We then feed the two acquired images to a conditional generative adversarial network (cGAN) to generate the phase image of a complex object. We show that the cGAN is able to directly learn the mapping relationship from the intensity pair to the targeted phase distribution. The effectiveness of the proposed approach is validated using both simulation and experimental data.