New Algorithm for Binary Connected-Component Labeling Based on Run-Length Encoding and Union-Find Sets

Based on detailed analysis of advantages and disadvantages of the existing connected-component labeling (CCL) algorithm, a new algorithm for binary connected components labeling based on run-length encoding (RLE) and union-find sets has been put forward. The new algorithm uses RLE as the basic processing unit, converts the label merging of connected RLE into sets grouping in accordance with equivalence relation, and uses the union-find sets which is the realization method of sets grouping to solve the label merging of connected RLE. And the label merging procedure has been optimized: the union operation has been modified by adding the "weighted rule" to avoid getting a degenerated-tree, and the "path compression" has been adopted when implementing the find operation, then the time complexity of label merging is O(nα(n)). The experiments show that the new algorithm can label the connected components of any shapes very quickly and exactly, save more memory, and facilitate the subsequent image analysis.