Cleaning up the Filled Clumps

The use of cellular automata to clean up the edges reduces the likelihood of ``holes'' within the clump edges, but does not eliminate this risk entirely. When the clump-filling algorithm described above encounters a hole in the edges surrounding a peak, the clump identifier value will ``leak out'' through the hole into the surrounding areas. This is where the limitations of the filling algorithm have a positive benefit, in that they prevent the leak from spreading round corners without limit. Instead, such leaks will tend to be produce straight features radiating out from a clump parallel to a pixel axis, which will terminate as soon as they meet another edge.

It is thus possible for the two or more clumps to ``claim'' a given pixel. This will happen if there are holes in the edges surrounding the peaks which allow the filling process to leak out. In this case, each pixel is assigned to the clump associated with the nearest peak.

Another cellular automata is used once the filling process has been completed to reduce the artifacts created by these leaks. This cellular automata replaces each clump identifier by the most commonly occurring clump identifier within a 3$\times$3$\times$3 cube (or 3$\times$3 square for 2D data) of neighbours. This process can be repeated to increase the degree of cleaning, by assigning a value greater than one to the configuration parameter Reinhold.FixClumpsIterations.

The results of this cleaning process are the final clump allocations for every data pixel, from which the catalogue of clump parameters is produced.