Shen et al. introduce a DBSCAN based algorithm for superpixel segmentation. The proposed algorithm consists of two steps: First, an initial superpixel segmentation is found using a spatially-restricted DBSCAN algorithm. Essentially, DBSCAN is augmented by a local search strategy. Second, the initial superpixels – which may be very small in regions with high color variation – are refined. To this end, superpixels whose size does into exceed a specific threshold are merged into the “closest” neighboring superpixels. This is repeated until all superpixel exceed the threshold. The results superpixels are illustrated in Figure 1 in comparison to other superpixel algorithms.
Figure 1: Qualitative results, from top to bottom: SLIC, LSC, ERS and DBSCAN.
What is your opinion on this article? Let me know your thoughts on Twitter @davidstutz92 or LinkedIn in/davidstutz92.
Shen et al. introduce a DBSCAN based algorithm for superpixel segmentation. The proposed algorithm consists of two steps: First, an initial superpixel segmentation is found using a spatially-restricted DBSCAN algorithm. Essentially, DBSCAN is augmented by a local search strategy. Second, the initial superpixels – which may be very small in regions with high color variation – are refined. To this end, superpixels whose size does into exceed a specific threshold are merged into the “closest” neighboring superpixels. This is repeated until all superpixel exceed the threshold. The results superpixels are illustrated in Figure 1 in comparison to other superpixel algorithms.
Figure 1: Qualitative results, from top to bottom: SLIC, LSC, ERS and DBSCAN.