An improved Directional Weighted Median Filter for restoration of images corrupted with high density impulse noise

This paper presents an approach to improve the performance of Directional Weighted Median (DWM) Filter based restoration of images corrupted with fixed valued impulse noise. The proposed approach involves minimum absolute difference criteria to distinguish among the edge and non-edge pixels. The identified corrupted pixels are then replaced by weighted median or mean value computed within the local window employing the proposed decision criteria. Further, to filter the residual noise at high density corruption; the proposed approach converges in less than five iterations. Simulation results show an improvement in visual quality of restored images both in terms of noise filtering as well as preservation of edges and structural content.