EPI-guided adaptive refocusing for high-SNR 3D measurement in a large-DOF MEMS-based structured light field

Structured light field (SLF) 3D measurement integrating micro-electro-mechanical system (MEMS) projection with unfocused plenoptic cameras (UPC) offers a promising solution for large depth of field (DOF) metrology. However, the high coherence of laser illumination and small UPC sub-apertures generates severe speckle and low signal-to-noise ratio (SNR), degrading reconstruction accuracy. To alleviate these issues, this paper proposes an EPI-guided adaptive refocusing model (EGARM). The model employs a phase-encoded epipolar plane image (EPI) analysis to extract phase features and adaptively estimate a depth-related slope factor. This factor subsequently guides digital refocusing to aggregate multiple sub-aperture views from the 4D light field. By averaging speckle patterns across viewpoints, the method synthesizes sharp images while effectively suppressing speckle. Additionally, a refocused ray calibration and global geometric refinement framework is introduced to impose global geometric constraints and help limit the degradation of reconstruction accuracy across the measurement volume. Experimental results demonstrate stable 3D reconstruction over an extended depth range of approximately 400-1856 mm, during which the effective field of view (FOV) expands from 120 mm × 200 mm to 580 mm × 780 mm, with reconstruction accuracy consistently better than 0.45 mm.