Removal of impulse noise from color image

This paper presents a vector median filter that includes a new mechanism for the detection of impulses in color images prior to further processing operations. The proposed filter has been tested for images corrupted by two sided fixed impulse noise model. If the central vector pixel in a kernel is found to be corrupted, it is replaced by the vector median of the kernel, else it is kept unchanged. Since the detection mechanism uses absolute deviation values, the computational complexity is reduced. The experimental results show a better performance of the proposed algorithm in lesser execution time as compared to a number of existing schemes for restoring color images corrupted by impulse noise.