运动模糊图像点扩散函数尺度鉴别

运动模糊图像的运动模糊方向被高精度地鉴别出来后,将之旋转到水平轴,则二维问题可简化为一维问题来处理。将旋转后的模糊图像在水平方向上进行一阶微分,然后在水平方向上求微分图像的自相关,并将自相关图像的各列加起来,得到一条鉴别曲线,在鉴别曲线上的零频尖峰两侧,会对称地出现一对共轭的负相关峰,两负相关峰间的距离等于运动模糊点扩散函数尺度的两倍。这一新的运动模糊点扩散函数尺度鉴别方法,不仅适用于匀速运动模糊尺度鉴别,也适用于振动、加速运动等变速直线运动的模糊尺度鉴别,而且具有很强的抗噪声干扰能力,鉴别精度高。文中借用信息光学中用于目标识别的光学联合变换相关器JTC概念,给出了该新鉴别方法的解析理论分析。

Identification the Scale of the Point Spread Function from the Motion Blurred Image

The direction of the motion which blurs the image can be identified with high precision.If it is rotated to the horizontal axis,the image restoration can be processed in1dimension instead of in2dimensions.After computing the first differential in the horizontal axis from the rotated blurred image,calculating the autocorrelation of every row of the differential image,pulsing the arrays of the autocorrelation image,an identify curve can be gained eventually.In the identify curve,there are two autocorrelation peaks,which point down with negative values.The position difference between the peaks is exactly double of the scale of the point spread function from the motion blurred image.This new identify method does work not only from the regularity motion blur image,but also from accelerated motion or vibration blur image.It can work out with high precise result even the motion blur image is dirtied with strong noises.The detail analysis is given.It is funny to find that there is an analogy between the identify result and the work out of Joint Transform Correlator(JTC),which is used for object recognizing in the domain of optic information processing.