As one of the most typical nonlinear filters

weighted median filter has considerable filter behavior by assigning weights. But each pixel has only one weight in the traditional weighted median filter. Correlations between pixel and other pixels under the filtering template are not considered. A novel weighted median filter model is proposed here where the weights correspond to these correlations. In order to keep stability when filtering on rotated and reversed images

symmetry of the correlations is utilized and can reduce the model complexity. Differential evolution algorithm is adopted for the weights optimization. Experiments about weights optimization show that this rotation invariant model exhibits faster convergence characteristics. Impulsive noise removal both on different gray images and color images show that it preserves better properties of edges or details

and has efficient noise attenuation and chromaticity retention.