Extending the Application of Conditional Signal Adaptive Median Filter to Impulsive Noise

Abstract

In this paper, two major improvements are applied to the Conditional Signal Adaptive Median (CSAM) filter to accommodate for high density impulsive noise with significant dynamic range. Homogeneity level of a pixel gives a good estimate in separating noisy pixels from pixels belonging to an edge or a fine detail in an image, but the proper choice of a threshold of homogeneity is highly affected by the noise density and range. A more robust method is developed for choosing such a threshold than in the original CSAM filter which extends the application of the filter to impulsive noise not only salt and pepper noise. Results are also compared for both low and high density impulsive noise.