基于相位相关的匀速直线运动模糊图像位移参数估计

为估算匀速直线运动模糊图像的运动参数,提出了一种基于相位相关分析的图像配准方法。该方法利用傅里叶变换的平移特性,对产生平移的目标图像进行傅里叶变换并计算归一化互功率谱,其傅里叶逆变换对应二维脉冲函数,通过计算脉冲函数峰值坐标获取位移图像之间的亚像元级位移量。结合相位相关配准原理和线性空间不变退化模型,给出了匀速直线运动点扩散函数及其光学传递函数的数学描述;讨论了匀速直线运动模糊对相位相关配准结果的影响,证明了图像经过匀速直线运动退化后,位移图像之间归一化互功率谱具有不变性。实验结果表明：动态运动模糊图像最大检测误差为0.489 pixel,标准差为0.16 pixel。

Estimation of displacement parameters for uniform linear motion-blurred images based on phase-only correlation

To estimate displacement parameters for uniform linear motion-blurred images,a method of image registration based on phase-only correlation is introduced.The method is based on the translation characteristics of Fourier transform to search the translation between the images.The inverse Fourier transform of the normalized cross power spectrum for a pairs of images is a two dimension impulse function,and the translation displacement between the images can be finely retrieved from the location of the correlation peak.Combined with the image registration principle of phase-only correlation and the degradation model of spatially variant blur,the mathematical expressions of Point Spread Function（PSF） and Optical Transfer Function（OTF） for the uniform linear motion blur are described,and the effect of motion blur on the correlation registration is discussed.The relation between the variance of normalized cross power spectrum and the degradation of blurred images is demonstrated,which is only dependent on the displacement between the images and not on the blur in images.The experimental result shows that this method can detect sub-pixel-level shift parameters accurately and efficiently and the maximum registration error is only 0.489 pixel,the root of mean square error of image registration is 0.16 pixel.