ABSTRACT
ABSTRACT: Image processing technology has driven the biomedical imaging to a profound change. Capable for a wide range of applications such as biomedical image processing, skin detection has developed rapidly with various approaches and methods of human skin segmentation. However, the principal obstacles that skin detection faces are still the different degrees of skin tone color, illumination conditions and skin color-like backgrounds. Given that, in this paper, we proposed a novel fusion strategy for dynamic skin detection, which is based on a smoothed 2-D histogram, Gaussian model and an online dynamic threshold calculated on face skin tone color, which reduces the training required, to a certain extent. In this approach, we adopted face detector to refine the skin model, as face is a prominent indicator of different characteristics of skin tone color, especially in images that include more than one face with different ethnicities. Qualitatively and quantitatively experimental results show that the proposed method is more robust and effective compared with state-of-the-art methods, because of its low computational costs and high accuracy.
