Malyshev A. Knyazev A.
Conjugate gradient acceleration of iterated edge-preserving smoothing filters
Reporter: Malyshev A.
We propose acceleration of iterated versions of edge-preserving smoothing filters by running them through a conjugate gradient method, without deteriorating their quality. The filter non-linearity is dealt with by careful restarting and freezing. Our initial numerical experiments demonstrate 20x acceleration of the classical bilateral filter and 3x acceleration of the recently developed guided filter.
To reports list