A Facial-Skin Condition Classification System in Wavelet Domain

This paper describes a facial-skin condition evaluation system based on cosmetician's suggestions to let the users realize their facial-skin conditions. Two kinds of features are extracted from the captured skin image. Four parameters of sub-band information of the gray part of a color skin image in wavelet domain are used as the first kind of features. The contrast, inverse difference moment, deviation information in the gray-value histogram, and the entropy of the co-occurrence matrix information are calculated as the second kind of features. The fuzzy rule extraction on the first-kind features is sued to distinguish the oily skin group from the others. The fuzzy c-means process with the second-kind features is used to classify the neutral skin and the dry skin. In the experiments, the classified results are compared with the cosmeticians' results, and the correct rates are 100% for the first test group of 30 test images and 92.8% for the second test group of 14 images.