A Deep Learning Approach for Noise Reduction of Off-axis Computer Generated Holograms

In this research, we propose a deep-learning-based computer generated hologram generation algorithm. The algorithm is able to generate a de-noised off-axis computer generated hologram. A de-noising convolutional neural network (DnCnn) is trained with added non-Gaussian and non-stationary speckle noise. Signal noise ratio (SNR), peak signal to noise ratio (PSNR), and mean square error (MSE) are used to evaluate the performance of the DnCnn. What's more, compared with the reconstructed image, the pixel distribution of the denoising image is closer to the original image. Results show that the algorithm is superior to conventional algorithms improvement about the quality of the reconstructed images.