| Documentation | Build Status | Code Coverage |
|---|---|---|
  |
 |
 |
This package allows to view an N-dimensional array as an 2N-dimensional TiledView being separated in overlapping tiles.
The tiled view has read and write access.
Even without explicitely using the TiledView datatype, you may be interested in using the function tiled_processingwhich is capable or automatically tiling the input array and reassembling the results (with overlap) via a weighted window approach. It can even deal with functions that return multiple results in a Tuple or Vector by returning a Vector of results. All you need to do to retrieve the assempled result is to apply the .parent member to each Tiled result.
Via the TiledWindowView it is possible to imprint a weight-window onto the tiled view. By default the window is chosen such that
it sums up to one except in places very close to the border, where an insufficient number of contributions are generated.
However this can effect can easily be accounted for, since it optionally returns an overall weight distribution.
Example:
julia> a = TiledView(reshape(1:49,(7,7)), (4, 4),(1, 1));
julia> size(a)
(4, 4, 3, 3)The toolbox also offers support for iterators on the tiles via the functions eachtile(), eachtilenumber(), and eachtilerelpos().
A very convenient way of processing all tiles with a user-supplied function and fusing the images automatically via window-based weighting is using
the function tiled_processing().
Type ]in the REPL to get to the package manager and install it:
julia> ] add TiledViews