Mean filter

The most common linear smoothing algorithm is the mean filter, and it is probably the most popular filter amongst interpreters. Typically, the weights in the kernel are uniform (Figure a & b), but they can also be triangular (inversely proportional to distance from the input sample, as in Figure c).

The output of a 5 × 5 uniform mean filter is shown here. Clearly, there is significant noise remaining in the horizon: this filter is not well suited to noise reduction. The problem is that noisy pixels—including anomalous spikes—are weighted the same as all the other pixels in the kernel. Starting with a despiking filter like the conservative smoothing filter might help, but it is easier to simply apply a different filter more suited to noise removal.

Of course, in some circumstances, this filter’s relatively weak effect could be an advantage, but in general Hall (2007) does not recommend its use.