Digital inline holographic reconstruction with learned sparsifying transform

Abstract We propose a digital inline holographic reconstruction method based on learned sparsifying transform. An iterative algorithm which includes the steps of updating background, phase, and image, as well as a step of sparse coding, is adopted to optimize the designed penalized least squares object function with regularization based on a sparsifying transform learned from sample images. A lensless inline holographic microscope (LIHM) was built and used to image a U.S. air force target and a pumpkin stem sample. The proposed method was applied to reconstruct the sample images. Compared with conventional holographic reconstruction method or the L1-norm-based penalized least squares reconstruction method, the imaging results show that the proposed method could better suppress the twin-image disturbance, staircase edges and block artifacts, thus enhance the reconstructed image quality.