Details

Autor(en) / Beteiligte

• Li, Zhongsheng
• Lou, Shuang

Titel

An Extracting and Labeling Algorithm for Connected Components in Images

Ist Teil von

• 2021 International Conference on Artificial Intelligence, Big Data and Algorithms (CAIBDA), 2021, p.210-213

Ort / Verlag

IEEE

Erscheinungsjahr

2021

Quelle

IEEE/IET Electronic Library

Beschreibungen/Notizen

• Image connected region labeling can effectively improve the efficiency of subsequent image analysis and processing tasks. Many achievements have been made, but there is no labeling algorithm for color images. Therefore, we propose a multi-class dynamic labeling(MCDL) algorithm. We construct the mathematical model of the MCDL algorithm first. And then, a series of new concepts, such as run, affecting scope and column coordinate set, are put forward. Based on these concepts, the judgment criterion on connection of two runs is defined with set operations. Next, a theorem is given and proofed to limit the number of provisional labels, and a class example is used to demonstrate operations of the MCDL algorithm. After that, the storage model of results is described. Finally, the implement of the MCDL algorithm is discussed. The MCDL algorithm is distinct from all conventional algorithms mainly on two aspects: it can label both binary images and color images, and these images may be irregular, but the conventional algorithms label only N×N or N×M binary images; it not only labels but also extracts the connected components in an image, whereas the conventional algorithms just label them. Comparisons and experimental results on various types of images show that the MCDL algorithm is efficient and robust.