MMBM-Driven and IMU-Assisted Adaptive Deblurring for Periodic Rotational-Scanning Panoramic Imaging Systems

A periodic rotational-scanning panoramic imaging system (PRS imaging system) can acquire large-scale, continuous, and high-resolution panoramic images through rotational scanning. However, non-ideal camera motion during exposure introduces spatially varying motion blur, which degrades image quality and affects subsequent visual interpretation. To address this problem in a self-developed PRS device, this paper proposes an adaptive image deblurring framework based on inertial measurement unit (IMU) assistance and the motion-based motion blur metric (MMBM). First, IMU data collected during exposure are used to calculate the MMBM, which represents the motion blur degree of the current image. The metric is then used to adaptively determine the iteration number of the Richardson–Lucy (R-L) deconvolution algorithm, avoiding unnecessary restoration and reducing artifacts caused by over-restoration. Second, the point spread function (PSF) size is adaptively determined from the camera motion trajectory, and the corresponding PSF is constructed to match different jitter intensities. Finally, an adaptive image partitioning strategy is introduced to handle spatially non-uniform blur caused by camera rotation. Experiments on real images collected by the self-developed dual-spectrum PRS imaging system show that the proposed method achieves stable restoration performance, preserves image naturalness, suppresses unnatural distortions, and reduces computational cost.