It is well known in the medical community that being overweight/obese is damaging to one's health, and increases the risk of diabetes. However, this is currently investigated via physiological/biochemical experiments considered to be invasive, or through the body mass index of a population. In this paper we propose a non-invasive method with statistical pattern recognition to analyze the relationship between three classes: Overweight/Obesity (140 samples), Healthy (125 samples), and Diabetes Mellitus (284 samples), using facial block color features. Facial images labeled by medical doctors practicing western medicine were captured by a specifically designed non-invasive device adopting color correction. After image capture, color features were extracted from the facial blocks. When comparing the three classes in groups of two by calculating their mean Euclidean distance, the results indicate Overweight/Obesity and Diabetes Mellitus are closer than the other combinations. This supports the claim that being overweight/obese is highly correlated with the presence of diabetes.
