Adaptive skin color tone detection with morphology-based model refinement

Abstract

Nowadays, many applications use the biometric feature to deal with human-computer interaction. The skin color detection is the one of the most common approaches to locate human body part such as face, hands, and arms. It plays an important role of the following processes. Most of the researches are focus on specifying the particular range of the color space and using morphological operator without any restriction. Practically, the range of skin color tone is dynamic changes depending on the lighting condition and shadow interference. Without restrictions, the result image is usually under- or over-estimated. Moreover, the difference of the ethnicities is usually results in the detection faults. This paper proposed an adaptive skin color tone detection using YCbCr color model. Based on the morphological closing and the constrained centroid-shifting algorithm to refine the skin color model. In experiments, it shows that our proposed method samples the target body part more compact and smooth.