Seeing Through Scattering With Computational Advances: A Review

Unlike clear media such as clear water and air, a scattering medium is inhomogeneous, in which propagating photons are multiply scattered and eventually randomized. The resultant light field is presented by numerous randomly distributed speckles that mix the information from both the incident optical field and the medium in either a linear or nonlinear manner. The access to the information before scattering can realize information transmission in a complex medium, biological imaging in a strong scattering regime and even optical encryption. Extracting information from the random speckles is therefore of technical concern, but that poses challenges. To address this, many solutions have been recently proposed, especially to quantify the transformation of the scattering process and its inversion, including analytical and learning‐based approaches. This review provides a concise summary of these advancements, which have enabled the mitigation of randomness induced by multiple scattering. These techniques have not only allowed for the theoretical and practical realization of “seeing through” scattering media across various optical dimensions but have even demonstrated the feasibility in kilometer‐scale applications, paving the way for broader adoption in scientific and industrial contexts.