The scalar median filter has been widely used
for reducing impulsive noise while preserving
signal edges in signals and images.
The vector median filter was introduced as a means
for filtering color images and other types of vector fields.
It has been stated that the vector median filter preserves
the color better than applying the scalar median filter to
each color component separately.
In this work, experiments and analysis were conducted
in order to quantify the relative performance
of the scalar median filter, the vector median filter,
and the extended vector median filter. The improvement in
the vector signal-to-noise ratio was used as a measure of
noise reduction. Based on the experimental results, the vector
median filter does not, in general, perform better than the
scalar median filter. The extended vector median filter performance
varies based on the variance of the image and the noise.
In general, the extended vector median performs better than the
vector median filter and similar to the scalar median filter.