Accelerated iterative blind deconvolution of still images

We propose an approach for blind deconvolution of still images with moderate noise contamination. The iterative technique is based on the general concepts of iterative techniques for blind deconvolution. The ill-convergence problem associated with most of the iterative techniques is circumvented in our approach using zero-sheet separation techniques. This technique can handle real images blurred by complex point spread functions (PSF), which is a common imaging problem, with blur signal-to-noise ratios (BSNR) of 70 dB and above for PSF of size 32/spl times/32 and larger. The technique performs much better for PSF of smaller sizes with low BSNR around 30 dB and provides convergence of the final solution with minimum iterations and is also capable of determining the size of the PSF.