运动物体成像的动态模糊仿真与恢复

Motion Blurring Image Simulation and Restoration in Dynamic Systems

摘要: 在对运动物体拍摄的过程中会出现图像模糊的现象，因此研究其降质的内在机理以及图像的恢复方法有着重要意义。本文通过构建一维和二维匀速直线运动模糊模型，揭示图像将质的内在机理，并根据其过程进行逆推导从而建立运动模糊图像恢复的数学物理模型。利用MATLAB进行仿真实验，模拟不同运动参数下的模糊成像与恢复过程。并通过主观视觉评估和客观指标评价，对比分析逆推导方法与现今普遍的方法。结果表明，本方法在特定运动场景下能有效恢复图像，并且在清晰度和细节保留上具有优势，为图像处理领域提供了有一定价值的参考。采用主观视觉评估和客观指标(峰值信噪比/结构相似性指数)进行的对比评估，实验结果证实了该方法在特定运动场景下的优越性能，与基准方法相比，在边缘保留方面提高了15%~20%，峰值信噪比提升了2.3分贝。所提出的框架在结构清晰度和细节保留方面均展现出优势，为光学系统优化和计算成像应用提供了有价值的见解。

Abstract: The inherent motion-induced image blur during dynamic object photography necessitates in-depth investigation into degradation mechanisms and restoration methodologies. This study establishes degradation models for uniform linear motion blur in both 1D and 2D domains, revealing the physical essence of image quality deterioration. Through inverse deduction of the blur formation process, we develop a mathematical-physical restoration framework. MATLAB-based simulations systematically demonstrate blur generation and recovery under varying motion parameters. The results show that this method can effectively restore images in specific motion scenes and has advantages in clarity and detail retention, providing a valuable reference for the field of image processing. Comparative evaluations employing subjective visual assessment and objective metrics (PSNR/SSIM) contrast our inverse-deduction approach with conventional Wiener filtering and Lucy-Richardson algorithms. Experimental results verify the method’s superior performance in specific motion scenarios, exhibiting 15%~20% improvement in edge preservation and 2.3 dB PSNR enhancement over benchmark methods. The proposed framework demonstrates advantages in both structural clarity and detail retention, providing valuable insights for optical system optimization and computational imaging applications.