Machine Learning (ML) has recently been applied to the problem of digital hologram generation with generally positive results due to the generation speed increase that is possible, because of the non-iterative inference step. In this work, a Convolutional Neural Network (CNN) is trained to generate holograms, and its performance is compared to state-of-the-art iterative methods, both in terms of reconstruction quality and computation time. Here, a CNN built on the UNet architecture, capable of hologram generation, is presented. The network is trained on 4096 images of varying spatial frequencies, both generated by hand and from the DIV2K dataset. It is compared to the most common iterative method for hologram generation, namely the Gerchberg-Saxton (GS) algorithm and its modern and improved implementations. In reconstruction quality, the neural network outperforms the original implementation of GS when evaluating Mean Square Error (MSE), geometric error (GE), Structural Similarity Index Measurement (SSIM), and Peak Signal-Noise Ratio (PSNR) of 64 unseen test images. However, on the same test images, the network lacks behind the modern, optimized GS implementations in all error and accuracy measurements. The network does, however, achieve these results at a rate 70-280 times faster than the iterative methods, depending on the particular implementation of the GS algorithm, which corresponds to a possible generation rate of the network of 32 FPS on average.
|Periode||22/01/2022 → 28/02/2022|
|Navn||Proceedings of SPIE, the International Society for Optical Engineering|