基于频谱边缘检测的运动模糊方向精确估计

针对运动模糊图像的模糊方向估计问题,详细分析了匀速直线运动模糊图像的退化模型,提出一种在频域精确估计运动模糊方向的方法。首先,计算退化图像的频谱,用高斯-拉普拉斯(LoG)边缘检测算子检测出频谱中的暗条纹轮廓;然后,用Radon变换找出垂直于暗条纹的角度;最后,根据图像长宽比确定频谱暗条纹和模糊方向之间的关系,计算出模糊方向。仿真结果表明,对模糊尺度从7到30像素的退化图像的模糊方向估计误差不超过1°,估计结果非常精确。

Accurate estimation of blurred motion direction based on edge detection of spectrum

With regard to the problem of estimating the blurred direction of motion-blurred image,the degradation model of blurred image in uniform linear motion was analyzed in detail,and a method which can estimate blurred motion direction accurately in frequency domain was proposed.Firstly,the spectrum of degraded images was calculated,Laplacian of Gaussian(LoG) edge detection operator was used to detect the contour of dark stripes in spectrum;then the Radon transform was employed to find the perpendicular angle to the dark stripes;finally,according to the aspect ratio of the image to determine the relationship between the spectrum dark stripes and blur direction,the blur direction was calculated.The simulation results show that estimated results are very accurate and the estimation error of blurred direction is no more than one degree when the blur scale of degraded image varies from 7 to 30 pixels.