Teaching of skin extraction algorithms for human face color images

Two methods for teaching of algorithms for the skin extraction in color images of human faces are proposed and discussed. The first method is based on self-organizing neural network called “growing neural gas”. The second one is based on morphological classification by minimal cutting of neighborhood graph for a training set in color space. The CIE Lab color space is applied for color description in both cases. The efficiency of both methods is demonstrated. The differences in selection results of the proposed methods are explored and demonstrated.