Tilt-corrected reflective ptychography based on automatic differentiation

Abstract

As a lensless computational imaging technology, reflective ptychography detects the reflected diffraction field of multiple overlapping areas and reconstruct the complex amplitude effectively. However, in reflective ptychography system, the existing of reflection angle will hamper the reconstruction results due to the reflective diffraction distortion and overlapping inconsistency, especially under large angle condition. In this paper, we interpret the transformation formula between tilted and non-tilted diffraction patterns as the combination of rotation and translation, and propose a tilt correction method based on automatic differentiation, eliminating the need for accurate prior knowledge about tilt angle and the need for interpolation operation. Specifically, this approach enhances the reconstruction quality under large angle reflection effectively. A series of simulations and experiments are conducted compared with traditional interpolation tilt correction method. The results indicate that our method will enable more accurate and effective correction of tilted diffraction patterns, thereby improving overall reflective ptychography reconstruction quality. The potential application of such a strategy can be expanded to other reflective imaging systems with large reflection angle, such as scenes measuring samples by using transparent substrate and simultaneously requiring large angle reflection to ensure high reflectivity.