Computational holography reconstruction method based on deep learning

The invention discloses a computer-generated hologram reconstruction method based on deep learning. The method comprises the following steps: S1, a training stage; and S2, a test stage. According to the computer-generated hologram reconstruction method based on deep learning, the computer-generated hologram image of the tumor cell is processed and analyzed by using a deep learning algorithm so as to generate a tumor cell hologram without interference of a zero-order image and a conjugate image, and reconstruction of the hologram is realized; according to the method, speckle noise is effectively inhibited, and the problem that zero-order images and conjugate images are difficult to remove is solved, so that the structure and phase information of tumor cells are accurately reproduced. 本发明公开了一种基于深度学习的计算全息重建方法，该方法包括以下步骤：S1、训练阶段；S2、测试阶段。本发明采用上述一种基于深度学习的计算全息重建方法，利用深度学习算法对肿瘤细胞计算全息图像进行处理和分析，以生成无零级像、共轭像干扰肿瘤细胞全息图，并实现对全息图的重建；该方法有效地抑制了散斑噪声，解决了零级像和共轭像难以去除的问题，从而精确再现了肿瘤细胞的结构和相位信息。