Locally adaptive space variant restoration of CT images

Summary form only given. An iterative algorithm for space variant image restoration was applied to computed tomographic (CT) images for improving spatial resolution compromised due to the finite size of the X-ray beam profiles. A prewhitening filter is designed from the noise power spectrum to uncorrelate the noise in the image. The restoration problem is formulated as a maximum-likelihood solution regularized with the weighted image norm. The nonstationary variance of the noise at each pixel location is utilized to locally regulate the smoothness or sharpness of the restored images. Unlike previous space variant restoration methods, this restoration procedure is carried out in a Cartesian coordinate system. Therefore, the symmetries in the spatial variations of the point spread function (PSF) are utilized to develop an efficient storage scheme which requires 0.75 megawords of memory for a 256*256 image, and a 9*9 PSF. This makes the computation of the reblurring process as fast as for a space invariant PSF.<>