一种改进的运动模糊图像修复参数计算方法

An Improved Method of Calculating Restoration Parameters for Motion Blurred Image

针对运动模糊图像修复时难以准确计算模糊方向和模糊长度2个参数的问题,文章提出了一种基于图像频谱分布特征的模糊图像修复参数计算的改进方法.考虑到传统的傅里叶变换导致图像边界产生频谱骤降的问题,设计了基于高斯滤波的图像边界模糊处理来消除频谱亮十字的改进方法,在此基础上,通过对频谱图像进行垂直方向线积分,对局部极小值求均值,从而准确求得运动模糊方向和模糊长度.最后以无人机航拍模糊图像为例,对改进的运动模糊参数计算方法进行了验证,并将其与相关研究成果进行对比,发现本方法求取的运动模糊长度与设定值相差约为0.06,而且模糊长度在10~30个像素时,模糊方向参数的检测质量有明显提升.

Focused on the difficulty to accurately calculate the parameters of fuzzy direction and fuzzy length for motion blurred image restoration, this paper presented an improved method for the calculation of fuzzy image restoration parameters based on the characteristics of image spectrum distribution. Considering the problem of sudden drop of image caused by traditional Fourier transform, the improved method of image boundary blur based on Gaussian filtering was designed to eliminate spectral bright cross. On this basis, the spectral image was taken in the vertical direction and the mean value of the local minimum was obtained, and the motion blur direction and the fuzzy length were obtained accurately. Finally, the fuzzy motion parameters of the UAV aerial image are validated and compared with the related research results. It is found that the motion blur length calculated by this method differs from the set value by about 0.06, and when the fuzzy length of 10 to 30 pixels, the detection quality of the fuzzy direction parameters has improved significantly.