Self-Guided Wiener Deconvolution for Annular-PSF Coded Aperture Imaging

Coded aperture imaging (CAI) is a widely used computational imaging technique for lensless indirect imaging. A typical CAI system consists of two stages: optical recording and computational reconstruction. During optical recording, the object information is encoded by a coded mask (CM), while during computational reconstruction, the recorded image is processed using the system point spread function and a computational reconstruction method (CRM). The imaging performance depends strongly on the choice of both the CM and the CRM, in addition to the imaging conditions. In this work, we investigate a CAI system employing an annular-focusing diffractive lens as the CM and Wiener deconvolution (WD) as the CRM under spatially incoherent illumination. A previously unreported pixel-dependent variation in the behavior of object and background pixels is observed as the Wiener regularization parameter varies. Exploiting this behavior, we propose a self-guided Wiener deconvolution (SG-WD) algorithm that extracts object signals while suppressing background noise. Both simulation and optical experiments demonstrate that SG-WD improves the reconstruction quality, achieving approximately 14% lower RMSE and 30% higher SSIM than conventional WD in experiments, while also outperforming several widely used reconstruction methods in RMSE. As the proposed SG-WD is implemented entirely as a post-processing algorithm without requiring any hardware modification or additional optical recording, it can be readily integrated into existing computational imaging systems, including coded aperture imaging, holography, and computational optical microscopy.