对运动模糊图像进行恢复处理需要两个关键参数的支持:模糊角度和模糊长度。为此,提出了一种时间复杂度比传统Radon变换更低的新方法对频域中的图像进行角度检测。根据傅里叶频谱的特性,新方法以扇形方式展开,扫描出一条穿过圆心且亮度最大的直线,该直线的倾斜角度为频域图像中平行条纹的倾斜角度;利用Radon变换的性质将先前求得的模糊角度进行投影求和得到模糊长度。实验结果表明,新方法不仅降低了时间复杂度,还提高了模糊角度和模糊长度检测的准确性。
Recovering the motion-blurred images requires two key parameters to support: blur direction and blur length. A new method is proposed to detect the image angle in the frequency domain with a lower the time complexity compared with the tradi- tional Radon transform. According to the characteristics of the Fourier spectrum, this method scans the brightest line which through the center of the circle and form fan shape. The inclination angle of the straight line is the required angle of the image in the frequency domain. After that the Radon transform is used to obtain the blur length making use of previously calculated the blur direction. The experiment shows that the method not only reduces the time complexity, but also improve the accuracy of the blur direction and blur length detection.