Programmable hardware system for class of nonlinear order statistics rational hybrid filters - Electronics Letters

Zntroduction: Nonlinear filters have been found effective in removing different types of noise from images. This is especially the case if the noise corrupting the image is not Gaussian andlor image details are not to be blurred. For this purpose, a number of order statistics based techniques have been proposed [ 11. In this Letter we describe an extension to our work in [2], where we first proposed the hybrid filter structure. Here, we present the results of a study into the performance of these new nonlinear rational type hybrid filters and describe a programmable hardware system for implementing them. Three order statistics subfdters are first computed in the first layer of the hybrid structure. The idea here is to remove impulsive type noise while causing minimal distortion in the image. Order statistics have been found to be very effective in such applications. The final layer of OSRH fdters is a rational function. The aim of the final layer, in addition to its detail preserving capability [3], is to attenuate Gaussian type noise. The hybrid approach is therefore suitable for applications where both impulsive and Gaussian noise are expected.