Wiener filter improvement on the sphere

摘要

The Wiener filter presents an important technique to remove degradation in images due to unfocussed optics or non ideal recording systems, i.e. restoration. The rough classical wiener filter application to the omnidirectional images leads to erroneous results; these ones usually used in autonomous navigation of robots, for its large field of view, consequently facilitates the perception in inaccessible environments for human being, and in automatic remote vision; present important anamorphoses, so that recently some scientists focused on the improvement of classical algorithms proceeding with plane images to adapt them with spherical ones; noticing the limits of such methods particularly no taking into account the geometry of the image, various solutions have been proposed, nevertheless, most of them were partly adapted. This work extends the deconvolution technique using optimal Wiener filtering and regularization for plane images; proposed by E. Sekko to the spherical ones on the equivalent sphere, relying on the tools as defined by Daniilidis, after suggesting a new three-dimensional suitable wiener filter design.