2-D nonlinear IIR-filters for image processing-an exploratory analysis

A novel nonlinear IIR (infinite impulse response) filter structure is introduced and its deterministic properties are analyzed. It is shown to be better suited for image processing applications than its linear shift-invariant counterpart. The novel structure is obtained from causality inversion of a 2-D quarterplane causal linear filter with respect to the two directions of propagation. It is demonstrated that, by using this design, a nonlinear 2-D lowpass filter can be constructed which is capable of effectively suppressing Gaussian or impulse noise without destroying important image information.<
>