Identifying the Clump Edges

The initial identification of the edges is done by considering a set of 1-dimensional profiles through the data array. Each such profile can be represented by a plot of data value against distance (in pixels) along the profile. For each such profile, the algorithm proceeds as follows.

1. Find the highest remaining (i.e. unused) data value in the profile.

2. If this value is less than a specified background level (given by the configuration parameter Reinhold.Noise), then there are no remaining significant peaks in the profile so continue with the next profile.

3. Work out away from the peak position along the profile in both directions to find the edges of the peak. A peak ends when it either i) meets a pixel which has already been included within another peak, or ii) two adjacent pixels are both below the background level, or iii) the average gradient of the profile over three adjacent pixel drops below a minimum value specified by the configuration parameter Reinhold.FlatSlope, or iv) the end of the profile is reached.

4. If the peak was not truncated by reaching either end of the profile, and if the peak spans sufficient pixels, the positions of the two edges of the peak are marked in a mask array which is the same shape and size as the 2D or 3D data array. The minimum number of pixels spanned by a peak in order for the peak to be usable is given by the configuration parameter Reinhold.MinPix.

5. The position of the peak itself is also marked so long as its peak value is above a specified minimum value (given by configuration parameter Reinhold.Thresh).

6. The pixels within the 1D profile which fall between the upper and lower edges of the peak are marked as ``used'', and the algorithm loops back to the start (i.e. step 1 above).

This algorithm is applied to each 1D profile in turn. These profiles are divided into groups of parallel profiles; the first group contains profiles which are parallel to the first pixel axis, the second group contains profiles which are parallel to the second pixel axis, etc. There are also groups which contain parallel profiles which proceed diagonally through the data array. Thus there is a group of parallel profiles for every pixel axis and for every possible diagonal direction. Within each group, there are sufficient profiles to ensure that every element in the data array is included in a profile within the group.

Once all profiles have been processed, a 2 or 3D array is available that identifies both the edges of the peaks and also the peak positions themselves. Pixels which are flagged as peaks are only retained if the pixel was found to be a peak in every profile group. That is, pixels which appear as a peak when profiled in one direction but not when profiled in another are discarded.