We present a new technique for blind restoration of images degraded by a smooth, spatially-invariant, zero-phase blur function. The restored image is obtained by using the information preseived in the phase of the blurred image to form an initial estimate. Successive estimates are produced by iteratively refining the initial estimate. For many images, only a few iterations are required to produce good quality deconvolved images. We provide several examples demonstrating the effectiveness of the new blind deconvolution approach and a discussion regarding the relative merits of the method.