Compute arbitrary function in sliding window

This function can compute any function which operates on column-oriented data over a sliding centered window on column-vector data. It is vectorized and thus trades memory-usage for speed.
For small window sizes this outperforms functions like SMOOTH. But it is not just a fast moving average. You can specify any function handle that operates on column-oriented data and produces a scalar output, e.g. median, std, prctile (with a scalar 'p'), range, min, sum. But you could also create your own function. For example the (quite obscure) function handle
@(X) all(bsxfun(@gt,X(ceil(size(X,1)/2),:),X(1:floor(size(X,1)/2),:)),1) &
all(bsxfun(@ge,X(ceil(size(X,1)/2),:),X(ceil(size(X,1)/2)+1:end,:)),1)
will return a logical array indicating local maxima over a given window size reporting only the first maximum at plateaus.