Thin lines filtering
A. Atta, A. Darwish, Mohamed K. Metwally · 2005
A proposed thin lines (TL) filtering method is addressed in order to remove impulse noise from image features that fill less than half the area of the transform window such as thin lines, bands, and corners. Validation of the proposed method has been carried out by considering groups of different synthetic and real images corrupted by impulse noise at different densities (due to the paper's limited space, only a single synthetic image is considered). The performance is measured numerically and visually. To support our method, a comparison with the CWM filter is used.