Adaptive-length median filters for image processing

The authors propose algorithms using adaptive-length median filters, for improving impulse-noise-removal performance for image processing. The algorithms can achieve significantly better image quality than regular (fixed-length) median filters when the images are corrupted by impulse noise. One of the advantages of the algorithms is that, when realized in hardware, the additional circuitry required is rather simple. Furthermore, the algorithms can easily be integrated into efficient hardware realizations for median filters. Performance-simulation results and implementation issues are discussed.<>