Restoration of Images with Piecewise Space-Variant Blur

摘要: We address the problem of space-variant image deblurring, where different parts are blurred by blur kernels. Assuming a region-wise space variant point spread function, we first solve for case known kernels and boundaries between regions in image. then generalize method to challenging unknown domains. Using variational level set techniques, is processed globally. The deconvolution process stabilized unified common regularizer, thus preserving discontinuities differently restored regions. In subregions unknown, segmentation procedure performed using an evolving guided edges derivatives.