Local Pixel Statistics and Correlation Based Impulse Detection for Denoising of Digital Color Images

In this paper we present a new impulse detection scheme to remove impulse noise from color images. The proposed impulse detector uses the filtering window statistics and the correlation among color components for identification of noisy pixels. The use of correlation existing among RGB channels further improves the reliability of the detector. The restoration of the noisy pixel is done on the basis of brightness and chromaticity information obtained from the neighbouring pixels in the filtering window. The proposed method performs better than both the component-wise filtering as well as vector-based filtering approaches. The efficacy of the present method is demonstrated through extensive computer simulations.