The median filter is a non-linear filter. Its principle is simple: each pixel is replaced by the
median value in a defined neighborhood. For instance, for a
3x3 pixels neighborhood, each color channel of each pixel is
replaced by the fifth value among the 9 of the neighborhood
sorted in increasing (or decreasing) order
The result of the median filter on a 3x3 neighborhood is illustrated below. The first
row gives the input noisy images, the second row gives the
results and the third and fourth rows show a detail
with colors which enhance contrast.
Gaussian (sigma=20) | Uniform (+/-20) | Salt and pepper (3%) | Periodic (jpg) |
Execution time on a AMD 1800+ for the 384x288 image: 0.04 seconds.
First remark: the filter gives excellent results on the
"salt and pepper" noise. However, other results are
poor. The periodic noise is still present.
Second remark: this filter keeps contours. But is has a
defect. It rounds corners (or any angle if less
than 180 degrees), as shown on a detail
below on the white labels.
|