Fast Parallel Connected Component Labeling Algorithms Using Cuda Based on 8-directional Label Selection

Connected component labeling (CCL) is a key step in image segmentation where foreground pixels are extracted and labeled. Sequential CCL is a computationally expensive operation and thus is often done within parallel processing framework to reduce execution time. Various parallel CCL methods have been proposed in the literature. Among them NSZ label equivalence (NSZ-LE) method seemed to perform best. In this paper we propose two new parallel CCL algorithms based on 8-directional label selection and show that they run 3 to 10 times faster than NSZ-LE depending on the characteristics of images.