Single image local blur identification

Abstract

International audienceWe present a new approach for spatially varying blur identification using a single image. Within each local patch in the image, the local blur is selected between a finite set of candidate PSFs by a maximum likelihood approach. We propose to work with a Generalized Likelihood to reduce the number of parameters and we use the Generalized Singular Value De- composition to limit the computing cost, while making proper image boundary hypotheses. The resulting method is fast and demonstrates good performance on simulated and real examples originating from applications such as motion blur identification and depth from defocus

Fast automatic myopic deconvolution of angiogram sequences

“…However, in our myopic context where both α and the width γ of the PSF are unknown, we observed that GML provides much better estimates of the tuning parameters than GCV. This confirms the conclusions of [3] who also favored GML rather than GCV to estimate the depth from defocus of real images.…”

“…which is a generalization of the criterion considered by Wahba [5] and Trouvé et al [3] as it takes into account non-i.i.d. noise and the a priori expectation of the parameters x = E(x | θ).…”

“…with u 2 W = u ·W· u and |W| + the product of the strictly positive eigenvalues of W. The weighting matrix W is also called the precision matrix by some authors [3]. If the covariance matrix of u is non-singular, taking W = Cov(u) −1 and noting that |W| + is then the determinant of W, the usual Gaussian distribution is retrieved.…”

“…(25), are very similar. Following Trouvé et al [3], we used the generalized singular value decomposition (GSVD) of the operators H and D 23), and GML parameter estimation. B: result following method in [11].…”

Theoretical performance model for single image depth from defocus

“…However, acquisition of several images requires the scene to remain still during the whole acquisition process, which is a strong hypothesis in practice, unless sophisticated imaging systems are used to separate the input beam [3,4]. More recently, Single Image DFD (SIDFD) alternatives have been proposed [5][6][7][8][9]. Acquisition is then simpler, while the processing becomes more complex, since only a single image of an unknown scene is available to infer the depth map.…”