Simple and efficient deviation correction framework for multi-height lensless imaging

The multi-height lensless imaging technique enables high-throughput imaging with a centimeter-scale field of view via full-frame sensor acquisition. However, minute errors in axial and radial displacements can cause sub-diffraction pattern shifts from their preset positions, compelling conventional approaches to rely on precision positioning devices and complex optical alignment procedures. To reduce system cost and enable rapid acquisition, this work is developed a computational lensless imaging framework integrating adaptive optical calibration and quantitative phase retrieval. The framework achieves automatic alignment through iterative optimization of sub-diffraction pattern spatial coordinates, while simultaneously minimizing the Structural Similarity Index between measured and computed intensities using a partitioned interval strategy coupled with stochastic gradient descent, thereby enabling adaptive calibration of axial displacement error. Simulation and experimental results indicate that, compared to conventional approaches, the proposed method achieves robust reconstruction across various specimens under coupled multi-parameter deviations in the system. This technique attains sub-pixel radial localization accuracy and keeps axial distance correction errors under 1.05%, with the added computational cost limited to only about 1.4× that of conventional methods. Even under extreme conditions relying solely on mechanical scales for rough positioning, high-quality complex-amplitude imaging can still be achieved. This demonstrates that the proposed approach can significantly simplifying system design, reducing hardware costs, and completely eliminating the reliance on demanding physical alignment processes and high-precision displacement devices.