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

Area of Science:

• Image Processing
• Computational Imaging
• Robotics and Automation

Background:

• Periodic rotational-scanning panoramic imaging systems (PRS) acquire high-resolution images but suffer from motion blur due to non-ideal camera movement.
• This motion blur degrades image quality and hinders accurate visual interpretation.

Purpose of the Study:

• To develop an adaptive image deblurring framework for a self-developed PRS device.
• To mitigate spatially varying motion blur and enhance image quality for better interpretation.

Main Methods:

• Utilizing inertial measurement unit (IMU) data to calculate a motion-based motion blur metric (MMBM).
• Adaptively determining Richardson-Lucy (R-L) deconvolution iterations and point spread function (PSF) size based on IMU data.
• Implementing an adaptive image partitioning strategy to address non-uniform blur.

Main Results:

• The proposed framework achieves stable restoration performance on real-world PRS images.
• The method effectively preserves image naturalness and suppresses unnatural distortions.
• Demonstrated reduction in computational cost compared to traditional methods.

Conclusions:

• The IMU-assisted adaptive deblurring framework significantly improves image quality in PRS systems.
• The adaptive strategies enhance restoration stability and reduce artifacts.
• This approach offers an efficient solution for deblurring large-scale panoramic images acquired with PRS devices.