Identification of defocus blur parameters and restoration of degraded images*

Wu Qing,Guo Ping

2004-01-01

Abstract:The problem of restoration of images blurred by uniform defocus is important in many applications. The solution proposed here identifies important parameters with which to characterize the point spread function (PSF) of the blur, given only the blurred image itself. We take advantage of the feature that there are a series of concentric circles in the Fourier spectrum domain of the defocused image, and the distance between two neighbor circles is in inverse proportion to the defocus radius R. In order to estimate the distance between circles, we use a coordinate transform to map the spectrum image into a new Cartesian coordinate system, in which the concentric circles turn out to be parallel lines. After detecting the distance between two neighbor lines, the blur parameter can be obtained according to the relationship of inverse proportion. When the PSF is estimated, a fast Kalman filter is adopted to restore the given image.