ICMOS focal-plane-stitching-based image shift compensation approach

Intensified CMOS (ICMOS) is a low-light night-vision detection and imaging device composed of an image intensifier and a CMOS sensor through adhesive coupling. To achieve low-light imaging for large-aperture cameras, stitching multiple ICMOS units into a larger focal plane array (FPA) is a feasible solution. However, ICMOS inherently exhibits image shift; ignoring this issue may lead to image information loss in stitched FPAs due to insufficient overlapping pixels. This paper analyzed the main factors causing image shift in ICMOS imaging firstly. Then, based on the imaging characteristics of ICMOS, the Canny operator was used to extract the contour features of the imaging region, and a three-point circular arc fitting method is employed to identify the center of the effective imaging area. Multiple sampling and average calculations were performed to improve positioning accuracy. Finally, linear FPA stitching based on the effective imaging center achieves an inter-unit overlapping precision better than 0.02 mm, with the number of overlapping pixels meeting the specified requirements and ensuring the imaging integrity of the stitched focal plane. This paper systematically addresses the inherent image shift in ICMOS, achieves precise identification of the effective imaging center through feature edge extraction and fitting, and ensures stitching accuracy via image shift compensation.