Two-dimensional adaptive filter based on t-distribution assumption and full-plane support

A new objective function for adapting a two-dimensional filter is proposed. The optimal coefficient of the adaptive filter is determined by minimizing the objective function, that is, assuming that the error signal is independently and identically t-distributed. Small portions of large amplitude error signals is assigned a small weighting function, while a large weighting function is used for large portions of small amplitude error signals. By doing so, the effect of the large amplitude error signal is reduced. We applied the proposed algorithm to enhance contaminated images. The simulation results show that we can suppress more impulsive noise by using the t-distribution with small degrees of freedom assumption than by applying large degrees of freedom assumption. The performance of both small and large /spl alpha/ assumption is similar when the input is a Gaussian noise contaminated image.