A No-reference Edge Preservinggenetic Weight Adaptation Filter for Impulse Noise Removal

摘要

Image sensors and communication channels introduce impulse noise in video broadcasting applications. Median filters and its variants are widely used to remove impulse noise. The drawbacks of them are blurring of edge details and low noise suppression. To preserve the sharp and meaningful information in the image, the filtering algorithms are required to have an intelligence incorporated in it. Genetic Algorithm (GA) emerged as a promising solution in this direction. This paper proposes a Genetic Weight Adaptation Filter (GWAF) that provides directional optimization of restored pixel in a mean square sense resulting in replacement of noisy pixels by near originals along its edge direction. Various parameters like mean absolute error (MAE), percentage of noise elimination, percentage of pixels spoiled show that the proposed recursive no-reference filter performs significantly better than order statistic and fuzzy filters.