Three elliptical skin color models are presented for skin color detection. The first one is to model the skin color cluster using a single ellipse ignoring the lightness dependency. It is simple and efficient, and the skin color detection accuracy may be adequate for many applications. In the second model, the skin color ellipse is adapted to different lightness levels to better fit the shape of the skin color cluster. The model is more complex to train, and the computation efficiency is lower, but the skin color detection accuracy is considerably higher. In the third method, an ellipsoid is trained to fit the skin color cluster. It is almost as simple to train as the first model, but the skin color detection accuracy is higher. Having skin color detection accuracy almost as high as the second model, this model is easier to train and may be more efficient in computation.