Wavelet-based propagation model for the light modulated by a digital micromirror device

Digital micromirror devices (DMDs) support fast light modulation. The deviation between the actual modulation and the theoretical expectation is not negligible for quantitative phase imaging, structured illumination microscopy, and optical neural network. However, the existing models rely on simple approximation and non-adjustable sampling and calculation, which results in inaccuracy in tracking the actual optical field propagation and a hindrance to the applications of DMD. Here, we propose a Wavelet-based Propagation Model (WPM) that decomposes the complex modulation process into a series of wavelet diffractions. We validate our model through modulating complex fields such as vortex beams and random speckles. The calculated error of WPM was reduced to half compared with the conventional method. WPM significantly improved the calculation accuracy of light modulation, especially for the light phase. Furthermore, we demonstrated the potential of WPM through an optical diffraction neural network, and the result showed that the exact calculation of light modulation could help optimize the training of the network.