Cleaning the Clump Edges

The initial identification of clumps edges results in a mask array in which each data pixel is marked as either an edge pixel or a peak pixel or neither. Usually, the edge pixels can be seen to follow the outline of the visible clumps, but will often be badly affected by noise in the data. For instance, there may be holes in the edges surrounding a peak, or spurious pixels may be have been marked as edge pixels. Before continuing, it is necessary to reduce the effect of this noise. This is done in two steps described below.

1. The edge regions were ``dilated'' (i.e. thickened) using a cellular automata algorithm which proceeds as follows: if a pixel is marked as an edge pixel, then all immediate neighbours of the pixel are also marked as edge pixels. Each pixel is considered to be the central pixel in a square of 3$\times$3 neighbouring pixels for 2D data, or the central pixel in a cube of 3$\times$3$\times$3 neighbouring pixels for 3D data.

2. The thickened edge regions were then ``eroded'' (i.e. made thinner) using another cellular automata algorithm which proceeds as follows. If the number of neighbouring edge pixels surrounding a central pixel was greater than a specified threshold value (given by the configuration parameter Reinhold.CAThresh), the central pixel would be marked as an edge pixel. If the number of neighbouring edge pixels was equal to or below this threshold, the central pixel would not be marked as an edge pixel. This transformation can be applied repeatedly to increase the amount of erosion by setting a value greater than one for the configuration parameter Reinhold.CAIterations.