为了有效减弱高动态环境对图像采集的干扰，利用模糊图像的频谱特性，提出了一种适合不同模糊尺度的运动模糊方向估计方法。该方法对比不同模糊尺度图像频谱中心亮条纹边缘特征的差别，通过计算其曲率半径实现不同模糊尺度图像的量化鉴别，在确定图像模糊尺度的大小之后，分别利用倒频谱、二次傅氏变换对模糊图像频谱中心亮线特征进行提取，考虑图像高宽比、中心十字亮线的不利影响，在二值化处理的基础上剔除十字亮线，通过亮线特征点直线拟合完成模糊方向估计。实验结果表明，该算法能够有效鉴别不同模糊尺度的运动模糊方向，估计精度较其他同类算法更高，相对误差率均小于0.02。

In order to remove the high-dynamic infection on image collection, a method based on the characteristic of Fourier spectrum was proposed to solve different-scale motion’s direction detection. This proposed method used the curvature radio to distinguish the difference between the small-scale or large-scale motion. After that, the blurred direction with cepstrum domain or twice Fourier transform can be estimated by different-scale motion. Considering the ratio of the image and cross bright line, a method based on mid-value as a threshold to optimize the image was proposed to acquire the direction of the blurred motion. Experimental results show that proposed method can distinguish the different-scale motion effectively and can improve the precision of motion direction or scale estimation. The robustness of estimation algorithm is improved in comparison with other algorithms, the relative error rate is less than 0.02.