Modified hybrid median filter with local preserving for removal of low density randomvalued impulse noise in images / Muhammad Sailuddin Darus ...[et al.]

Random-valued impulse noise has proven to be more
difficult to remove from corrupted images than standard
impulsive noise due to its random nature. Because standard
impulse noise filters target the whole images pixels, useful image
details and information may be loss after the filtering process. To
filter the corrupted pixels and preserve important image details
during filtering, it is vital to introduce local preserving to the
filter. In this paper, a new filter for the removal of randomvalued impulse noise in digital grayscale images is presented. The
filter is made up of two different stages which is noise detection
and filtering steps, in which the properties of the random-valued
impulse noise and a modified hybrid median filter is used. This
noise detection is based on the histogram information of the
image corrupted by random-valued impulse noise to prevent the
filtering of noise free image pixels. The filtering of the detected
noisy pixels is performed afterwards using the modified hybrid
median filter. By introducing a local preserving scheme to the
modified hybrid median filter, the results obtained is shown to
outperform the standard median filter as well as the modified
hybrid median filter.