Single Shot Phase Retrieval

Single-shot phase retrieval aims to reconstruct a complex-valued signal from a single intensity measurement, bypassing the need for iterative algorithms and multiple measurements. Current research focuses on deep learning approaches, employing architectures like physics-driven multi-scale neural networks and those incorporating self-attention mechanisms and multi-layer perceptrons to improve feature extraction and reduce reconstruction errors. These advancements are significant because they enable faster, more robust phase retrieval in applications like coherent diffractive imaging, potentially revolutionizing fields requiring low-dose or high-speed imaging of radiation-sensitive samples.