对运动模糊图象的恢复而言,虽然维纳滤波法可解决反滤波法中H(u,v)零点的噪声放大问题,但无法消除由于模糊图象信息的不完整性而造成的边缘误差,该文所述的带最优窗的维纳滤波方法却可有效地去除噪声和减小边缘误差,除了边缘的L-形条带外,可得到近乎完美的恢复效果.通过对女孩和大狒狒头模糊图象的恢复试验证明,该方法效果较好.由于点扩展函数的设置是影响恢复结果好坏的关键问题,因此文中还证明了对任意方向的直线运动模糊图象的恢复只能利用直接设置二维点扩展函数来进行二维恢复,不可先进行x方向的一维恢复,再进行y

The paper discusses the restoration method for motion-blurred images. The large errors observed in inverse filter or Wiener filter restorations of images are mainly due to the fact that a truncated region of image data is available for processing. Wiener filter is less sensitive to noise than inverse filter for the restorations of motionblurred images. However, there is a large error component, called the edge error, that arises due to the fact that real images seldom have the periodicity assumed by discrete Fourier transform operation. Optimal windows for image restoration are designed on the basis of mathematical expressions for the restoration errors. With these optimal windows, near-perfect restorations can be obtained except the narrow L-shape border at the right and bottom edges of the image, which we have to sacrifice for reduction of restoration errors over the rest of the image. The images that vary gradually in intensity near their borders have the best results, as demonstrated in actual restorations of test images. The paper also proves that we can only directly setup two dimensional point spread function, instead of first restoring in x direction and then in y direction. A procedure for estimating point spread functions of real-world motion-blurred images is provided in this paper.