The advanced image blind restoration method is mainly reflected in the accuracy and rapidity of kernel estimation. Aiming at the problems of inaccurate blur kernel estimation and high time complexity caused by the redundant information or insufficient effective information in the current deblur methods, a blur kernel region estimation and space-variant restoration based on weighted $ {L}_{1} $ norm measure was presented. First, the multi-scale morphological gradient of the gradient image was extracted to suppress the interference of noise on the image; Then, the gradient weighted $ {L}_{1} $ norm measure was defined to be conducive to blur kernel estimation, the inaccuracy of blur kernel transformation caused by flat regions and tiny structure regions was solved, and the region of blur kernel estimation was obtained; Finally, the similarity of two or more regional blur kernels were used to determine the blur kernel estimation area of a space-invariant or space-variant degraded image. Since the selected kernel estimation region is much smaller than the whole image, the kernel estimation can be performed quickly. In the deconvolution phase, FFTW was used to do the calculation of Fourier transform, which greatly improved the speed of restoration. Extensive experiments show that proposed method can restore degraded image quickly and effectively.