| dc.description.abstract | Optical imaging through scattering media is a long-standing challenge with important applications ranging from observations through the turbulent atmosphere to imaging inside living tissues. Low-order aberrations due to random fluctuations of refractive index can be overcome by adaptive optics and turbulence-free ghost imaging; while the problem becomes intractable for optically opaque media in which strong light scattering scrambles the spatial information conveyed by light fields. Inspired by the heterodyne detection of the beat signal of two lasers, we propose a computational imaging scheme based on time-domain information encoding and fast-Fourier-transform-based information decoding that can realize non-invasive imaging through scattering media. The feasibility of the original idea is tested by a preliminary experiment that realizes imaging through scattering media by extracting the beat frequency of two lasers. The technique is further improved by replacing the beat signal of two lasers with an intensity-modulated laser. Using cross-spectrum detection of the modulation frequency and raster-scan measurement, we demonstrate that the image of an object can be reconstructed not only through both static and dynamic diffusers but also under extremely noisy environments, i.e., the light intensity is much lower than detector noise. To overcome the speed limit due to raster scan, the computational imaging mechanism is further improved to realize full-field imaging via space-time encoded pattern (STEP) illumination. We show that the images of objects can be reconstructed from a 1D time series of light intensity measured by a single-pixel photodetector. As a proof of concept, we experimentally demonstrate our technique with ground glass diffusers and slices of chicken breast as the scattering media. Various aspects of this technique, including resolution, penetration depth, imaging speed, and algorithm complexity, are discussed. | |
