Iterated Sor Filter Restoration of Colour Images
J.F. Saliba, Alan L. Harvey · Information Sciences, Signal Processing and their Applications · 1996
Colour images using a 24 bit RGB format for which 80% of the image data has been displaced by impulse noise can be restored by using an iterated convolution based filter which uses correlation between image values to restore each RGB colour component separately. This paper describes how the Successive Over Relaxation (SOR) filter does in fact give good results if individual colour components in RGB format are restored rather than the colour indexes [3]. Pictures degraded during transmission which have lost up to 80% of their data can be restored using this linear filter. This method does not require that the colour range of the image corruption or non-data region be explicitly known, only that it is evident from the corrupted image which extreme value is required for the filter output. Since the output produce new image colour values, the method requires some post processing using order statistic restoration methods.