An Improved Filtering Algorithm for Impulse Noise with High Density Based on Edge Information Preservation
Wei Zhao, Hui Li Jing, Bao Zhen Yang, Hui Zhang · Applied Mechanics and Materials · 2014
Aiming at edge information easily lost in impulse noise filtering, an algorithm for high noise density is proposed in this paper. Firstly, pixels contaminated can be detected by the characteristic of their grey and some information about every pixel and its neighborhood can be got. Secondly, by a judgment of the detail and a selection of the optimal neighborhood, two traditional filter methods are used to reduce the noise density and obtain more information of the contaminated pixels. Finally, the edge judgment based on double-window is added to preserve more edge information. Experiment shows that for impulse noise, whether the low density or the high density, the proposed algorithm can obtain more satisfactory result with a lower RMSE (Root of Mean Square Error) and a higher EPI (Edge Preserved Index).