Local Connected Component Labeling (CCL)

The identification of connected components can be a simple CCL method, the ICL method [Hendrickson2020], which uses both cell values and calculated normal vectors, the single pass threshold method described by [Chan2021], or a mix. Within one processor, the structure used to track connected components is an equivalent label set (ELS), as described in [He2007].

This program does not currently support multiple thresholds (ICL-M from [Hendrickson2020]).

Parallel Computations

The parallel process is broken into five phases, the first four based on [Harrison2011] :

1. Identify locally unique blob IDs on each processor - referred to as localID
2. Create globally unique identifications - referred to as globalID
3. Identify connections across processors
4. Solve cross-processor connections to determine globally connected blobs
5. Gather and calculate group properties for each blob and output

Solving the global connections in step 4 is done using the UFM method described in [Iverson2015].