Multichannel Wiener filters in color image restoration based on AR color image modelling

Color images are multichannel signals, whereas monochrome images are single-channel signals. Therefore, the extension of techniques used in monochrome digital restoration to color images is not straightforward. The interchannel correlations which exist in color images must be taken into account in the restoration procedure. The design of a multichannel Wiener filter in the frequency domain is presented. The sensitivity of the multichannel Wiener filter to the spectral estimates requires the use of a robust multichannel power spectral estimator. The use of multichannel autoregressive (AR) spectral estimators results in multichannel filters which have excellent performance.<<ETX>>