Visualizing an optically transparent object is typically achieved by retrieving its phase information that gets lost when one records the intensity of the light in a conventional microscope. Among various methods of phase retrieval, differential phase contrast microscopy is a technique that uses patterned illumination of the object and recording multiple images and post processing of these images with a reconstruction algorithm to obtain the complex field information of the object. Authors of this article have used an untrained neural network model that represents nonlinear image formation along with aberrations to compute the differential phase contrast images. This approach differs from conventional methods that rely on a linearized image formation model. The main advantage of this novel method is its applicability in imaging thick transparent objects with significant phase variations while effectively reducing aberrations.
You must log in
to add comments.