Bioinformatics and Biomedical Engineering: New Advances

176

Table 1. Comparison between our proposed fusion and non-fusion approach

in HSV color space using web images.

Classifier Fusion 2D-Histogram GMM

Dynamic

threshold

Accuracy 0.9057 0.8854 0.858 0.8914

F-score 0.6512 0.6189 0.6145 0.629

True positive rate 0.6521 0.6695 0.8014 0.8377

False positive rate 0.0679 0.085 0.1156 0.1039

as skin in the performance of the method of Tan & Chan (2012), while many skin pixels as

non-skin in the results of the approach of Ibrahim et al. (2012). It can be obviously observed

from Figure 3 that certain image samples, which are full of rather complicated conclusions

such as high illumination at row 5, skin color-lik ...