Row-Wise Corrective Median Filtering for Fast Impulse Noise Reduction in Color Images

Abstract

The vector median filter and its variants are popularly used to reduce impulse noise in digital color images. These methods work by the principle of similarity detection within a local block of pixels, termed a window and by replacing the centre test pixel by another pixel in that window that minimises the aggregate of a suitably chosen distance measure. This processing is performed over all the pixels in the image in a sequential fashion. This renders these filtering methods computationally expensive. In this paper, we propose a fast method for reducing impulse noise in color images. The key idea here is to slice each row of the image as a univariate data vector, identify anomalies (impulse noise) using anomaly detection approach and then apply a corrective median filtering technique over these to restore the original image. This idea ensures fast filtering as the iterations are now limited to the rows. Using simulations, we show that the proposed method scales efficiently with respect to a combined measure of accuracy and time.