Separable-denominator two-dimensional adaptive filters with application to noise reduction in images (Q1892058)

The paper approaches the realization of the separable-denominator two- dimensional (SD-2D) adaptive filters and their application to restoring images degraded by noise. Both the lattice-form and the direct form of these filters are considered. It is shown that the stability tests of SD- 2D filters are much simpler than those of general 2D filters. 2D least mean squares algorithms for updating the filter coefficients are given. The computational complexity of the SD-2D adaptive filters is analyzed in terms of the number of multiplications required for filtering and adaptation. Experimental results demonstrate that the SD-2D adaptive filters are superior to the nonadaptive SD-2D filters.