Simultaneous estimation of PSF and global frame shifts at lower S/N ratios

We present new algorithm for simultaneous estimation of degradation PSF and global frame shifts, using degraded frames only. Algorithm is based on the singular value decomposition (SVD). Separable approximation is supposed for the point spread function (PSF). It has been shown that frame shifts cause same shifts in the corresponding singular vectors. This fact is used in the spatial domain for PSF estimation, and in the temporal domain for estimation of global frame shifts. First few singular vectors contain less noise. This fact is further improved by an averaging operation. Magnitude functions of PSF first singular vectors are estimated by averaging of corresponding first few singular vectors of the degraded image. Estimation of global frame shifts uses corresponding first few singular vectors of consecutive frames. Experimental results on simulated and real-world degraded frames prove the validity of the approach.