基于Radon变换的运动模糊方向估计的改进算法研究

运动模糊图像的复原在天文，军事，工业控制，道路监控和刑侦方面具有重要的现实意义。在研究运动模糊图像复原中，对点扩散函数PSF（point spread function）的估计是关键点也是难点。通过对运动模糊图像进行傅里叶变化，得到其明暗相间条纹的频谱图。基于亮条纹的方向和运动模糊方向垂直的原理，利用Radon变化求解频谱图中亮条纹的方向，从而得到PSF中的运动模糊方向。由于模糊图像频谱图中十字亮线的存在，直接影响了实验结果的准确性,因此对于十字亮线产生的原因进行了分析，提出一种简单有效的去除十字亮线干扰的算法。实验结果表明：该算法简单、有效，能准确测量出模糊方向，误差控制在1°之内。

Study on improved algorithm for estimation of motion blur direction based on radon transform

Restoration of motion blurred images has practical significance in astronomy, military affairs, industrial control, road monitoring and criminal investigation. Estimation of PSF (Point Spread Function) is a key and difficulty in the study for restoration of motion blurred images. This paper carried out Fourier transform for motion blurred images to obtain a frequency spectrum diagram with bright and dark stripes. Based on the direction of bright stripes and the principle that the motion blur direction is vertical, the direction of the bright stripes in the frequency spectrum diagram could be obtained through Radon transform, thus the motion blur direction in PSF could be obtained. The cross-shaped bright line in the frequency spectrum diagram of blurred image could directly affect the accuracy of the experiment results. This paper analyzed the cause of the cross-shaped bright line, and put forward a simple and effective algorithm to remove the interference of the cross-shaped bright line. The experimental results indicate that the proposed algorithm is simple and effective, and it could accurately determine the blur direction through measurement with the error less than 1°.