Variations on the rank-order filtering theme for grey-tone and binary image enhancement

The author presents an adaptive rank-order filtering approach for developing a variety of image enhancement operators. The main idea is to choose a rank N(1⩽N⩽U=number of pixels in the processing window) depends on the local degree of match between the image data and an ideal model of the unimpaired grey value function. The aim is to develop filters that selectively act as maximum operators for the useful signal and as minimum operators for the noise. These perform better in many cases than median filters and can be used for both grey-tone and binary images. Fixed-rank filters can also be advantageously applied to binary images. They can be considered as generalizations of the well-known erosion, dilation, opening and closing operators of the mathematical morphology