Median filters with adaptive length

Two algorithms using adaptive-length median filters are proposed 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 algorithms, when realized in hardware, requires rather simple additional circuitry. Both algorithms can easily be integrated into efficient hardware realizations for median filters. The performance of the proposed filters is compared with regular median filters, generalized mean filters, and nonlinear mean filters. The hardware complexities of the filters are also compared.<
>