Efficient impulse noise removal for highly degraded images

46 p.In this thesis a two-stage noise removal scheme is proposed to process impulse noises from highly degraded images, comprising noise detecting and noise filtering. The noise detector first uses intensity difference within a subimage to cluster the pixels into two groups, the outliers and the fine pixels. Then a final detecting algorithm based on directional intensity difference is applied to the outliers to distinguish the true noise from the image details. The positions of the noisy pixels found by the detector are recorded as a noise map. Based on the noise map, a modified weighted mean filter is applied to remove the noise.Master of Science (Signal Processing