Improved Unsupervised Color Segmentation Using a Modified Color Model and a Bagging Procedure in -Means++ Algorithm

Table 10

Segmentation performance results for image “e” using GMM.

GMM	Cluster 1	Cluster 2	Cluster 3	Cluster 4	Metrics
					Best BAcc.: 0.98543; Mean BAcc: 0.87761; Mean sen.: 0.56276; Mean spe.: 0.94142; Avg. time: 4.03769
n					Best BAcc.: 0.99400; Mean BAcc: 0.95521; Mean sen.: 0.89530; Mean spe.: 0.96174; Avg. time: 3.09780
					Best BAcc.: 0.97908; Mean BAcc: 0.86823; Mean sen.: 0.65741; Mean spe.: 0.92706; Avg. time: 4.4489
					Best BAcc.: 0.99123; Mean BAcc: 0.90806; Mean sen.: 0.79366; Mean spe.: 0.94410; Avg. time: 1.9395
YCrCb					Best BAcc.: 0.98606; Mean BAcc: 0.86326; Mean sen.: 0.68289; Mean spe.: 0.92291; Avg. time: 2.966
CrCb					Best BAcc.: 0.99088; Mean BAcc: 0.90607; Mean sen.: 0.84163; Mean spe.: 0.93257; Avg. time: 2.29639
					Best BAcc.: 0.99261; Mean BAcc: 0.89050; Mean sen.: 0.69485; Mean spe.: 0.93121; Avg. time: 2.5547
					Best BAcc.: 0.99259; Mean BAcc: 0.93409; Mean sen.: 0.81547; Mean spe.: 0.94257; Avg. time: 2.22500
Orig					Best BAcc.: 0.99252; Mean BAcc: 0.85474; Mean sen.: 0.63664; Mean spe.: 0.86330; Avg. time: 2.3193