Process matrix in sliding blocks with user-supplied function.
Executes the function fun on each sliding block of size block_size, taken from the matrix A. Both the matrix A, and the block can have any number of dimensions. This function is specially useful to perform sliding/moving window functions such as moving average.
The output will have the same dimensions A, each one of its values
corresponding to the processing of a block centered at the same
coordinates in A, with A being padded with zeros for the
borders (see below for indexed images). In case any side of the block
is of even length, the center is considered at indices
floor ([block_size/2] + 1).
The argument func must be a function handle that takes matrices of
size block_size as input and returns a single scalar. Any extra
input arguments to nlfilter are passed to func after the
block matrix.
If A is an indexed image, the second argument should be the
string "indexed" so that any required padding is done correctly
as done by im2col.
Note: if func is a column compression function, i.e., it acts
along a column to return a single value, consider using colfilt
which usually performs faster. If func makes use of the colon
operator to select all elements in the block, e.g., if func looks
anything like @(x) sum (x(:)), it is a good indication that
colfilt should be used. In addition, many sliding block operations
have their own specific implementations (see help text of colfilt
for a list).
See also: blockproc, col2im, colfilt, im2col.
The following code
## creates a "wide" diagonal (although it can be performed more ## efficiently with "imdilate (A, true (3))") nlfilter (eye (10), [3 3], @(x) any (x(:) > 0))
Produces the following output
ans = 1 1 1 0 0 0 0 0 0 0 1 1 1 1 0 0 0 0 0 0 1 1 1 1 1 0 0 0 0 0 0 1 1 1 1 1 0 0 0 0 0 0 1 1 1 1 1 0 0 0 0 0 0 1 1 1 1 1 0 0 0 0 0 0 1 1 1 1 1 0 0 0 0 0 0 1 1 1 1 1 0 0 0 0 0 0 1 1 1 1 0 0 0 0 0 0 0 1 1 1
Package: image