Image enhancement in lensless inline holographic microscope by inter-modality learning with denoising convolutional neural network

Abstract

Compared to traditional optical microscope, lensless inline holographic microscope (LIHM) is more compact and low-cost. However, its resolution and imaging contrast are generally inferior mainly because of the twin-image background. In this paper we propose a deep learning-based approach to reduce the noise and enhance the imaging quality in LIHM by inter-modality learning from the traditional microscope images. By exploiting the denoising model in the learning processing, our network can be trained with a dataset synthesized from the direct-reconstructed images of LIHM and the high-resolution ground truth images obtained with a microscope. In the imaging process, other direct-reconstructed images of LIHM can then be enhanced by the trained denoising network. The image enhancement capability of our method was demonstrated by experiments with a U.S. Air force (USAF) target and a pumpkin stem sample. The results show that both the resolution and imaging contrast were significantly improved compared with traditional reconstruction methods in LIHM.