Computational terahertz phase imaging using a sequence of multi-plane intensity measurements

Advancements in terahertz (THz) technology have substantially propelled the capabilities of imaging systems; however, retrieving the complex phase distribution of arbitrary test objects within the THz spectral range continues to face fundamental and practical challenges. In this context, interferometry-based imaging schemes have been widely adopted for THz phase imaging. However, the efficiency of such techniques is fundamentally constrained by their reliance on phase-stable reference beams, rendering them particularly susceptible to noise and inherent imaging system instabilities. In this work, we address such limitations and experimentally investigate two non-interferometric imaging configurations integrated with a computational phase retrieval framework, enabling accurate recovery of complex-valued wavefronts from intensity-only data. The first configuration employs a lens-less imaging approach based on recording axial diffraction patterns at multiple planes. The second configuration captures defocused intensities near the focal plane of a 4f imaging system. Our proposed computational framework formulates the phase retrieval task as a nonlinear least-squares error optimization problem. Specifically, we employ an accelerated second-order optimization strategy by iteratively computing both the gradient and the Hessian of the cost function, which outperforms previously demonstrated multi-plane phase reconstruction approaches. Our imaging framework marks a significant step forward in computational THz imaging, with promising implications for diverse applications, such as non-destructive testing for materials characterization, biomedical diagnostics, and security screening.