filterN

Non-linear filtering of N-dimensional arrays. A function is applied, convolution style, to sets of elements defined by a mask.

Y = filterN(X, OP, MASK) performs N-D filtering of array X. The region defined by MASK is scanned across X to produce, at each step, the inputs to OP, whose output is placed in the corresponding element of Y.

OP may be a function handle or one of the strings "max", "min", "maxabs" or "sumsqr".

MASK may be a scalar cell to define a (hyper-)rectangular region, or a logical array to define a possibly scattered set of elements.

Optional arguments allow the form of the argument to OP and the behaviour at the boundaries of X to be specified. The output may be sampled onto a regular grid, avoiding unnecessary computation.

filterN is related to the Image Processing Toolbox functions nlfilter and colfilt.