White light three-dimensional imaging using a quasi-random lens.

Abstract

Coded aperture imaging (CAI) technology is a rapidly evolving indirect imaging method with extraordinary potential. In recent years, CAI based on chaotic optical waves have been shown to exhibit multidimensional, multispectral, and multimodal imaging capabilities with a signal to noise ratio approaching the range of lens based direct imagers. However, most of the earlier studies used only narrow band illumination. In this study, CAI based on chaotic optical waves is investigated for white light illumination. A numerical study was carried out using scalar diffraction formulation and correlation optics and the lateral and axial resolving power for different spectral width were compared. A binary diffractive quasi-random lens was fabricated using electron beam lithography and the lateral and axial point spread holograms are recorded for white light. Three-dimensional imaging was demonstrated using thick objects consisting of two planes. An integrated sequence of signal processing tools such as non-linear filter, low-pass filter, median filter and correlation filter were applied to reconstruct images with an improved signal to noise ratio. A denoising deep learning neural network (DLNN) was trained using synthetic noisy images generated by the convolution of recorded point spread functions with the virtual object functions under a wide range of aberrations and noises. The trained DLNN was found to reduce further the reconstruction noises.