Filling the Clump Edges

Once the edges have been cleaned up, the volume contained within the edges can be filled with an integer which identifies the associated peak. This algorithm proceeds as follows.

1. The mask array is scanned for pixels which are marked as peaks. Recall that only those pixels which are seen to be peaks when profiled in all directions have been retained. Each of these pixels thus represents a local maximum in the data value, and has a significantly higher data value than any of the surrounding pixels. Each such peak is given a unique integer identifier. This identifier is used within the following steps to label all pixels in the clump of emission surrounding the peak.

2. A line of pixels parallel to the first pixel axis, and which passes through the peak, is then considered. The line is followed away from the peak, in both directions, until pixels are encountered which are flagged as edge pixels. All the pixels along this line between the two edge pixels are assigned the clump identifier associated with the central peak.

3. For each such pixel, another line of pixels parallel to the second pixel axis and passing through the pixel is considered. The line is followed away from the pixel, in both directions, until edge pixels are encountered. All the pixels along this line between the two edge pixels are also assigned the clump identifier associated with the central peak.

4. For each such pixel, another line of pixels parallel to the third pixel axis and passing through the pixel is considered. The line is followed away from the pixel, in both directions, until edge pixels are encountered. All the pixels along this line between the two edge pixels are also assigned the clump identifier associated with the central peak.

The above process will fill the volume between the edges, but may miss some pixels (e.g. if the initial line parallel to the first pixel axis spans the clump at an unusually narrow point). In order to alleviate this effect, the above process is repeated, but scanning the pixels axes in a different order (2,3,1 instead of 1,2,3). For 3D data, it is repeated a third time using the axis order (3,1,2).

Even so, it should be recognised that this filling algorithm will fail to fill certain shapes entirely. For instance, ``S''-shaped clumps could not be filled completely using this algorithm.