A line labeling and region growing based algorithm for binary image connected component labeling

Abstract

We propose a connected component labeling algorithm using line labeling and region growing method (LRGM) in this paper. First, we analyse the basic characteristic of current labeling algorithms, and set the scan order of LRGM from left to right, top to bottom, to assign a label to all connected components. Second, we eliminate label conflict by region growing method, because a large number of K label arises of which many are equivalent. Finally, we optimize the search and judgment criterion of LRGM, to make the new algorithm is independent of connected components shape, and the search time is much less than it before costed. Experimenting on various types of document images (pictures, newspapers, etc.), we find that our method outperforms the other sequential methods issued in publication. It greatly increases the run efficient, and it is very useful for real-time and large images processing.