After the scene has been modeled, Pfinder watches for large deviations from this model. New pixel values are compared to the known scene by measuring their Mahalanobis distance in color space from the class at the appropriate location in the scene model, as per Equation 5.
If a changed region of the image is found that is of sufficient size to rule out unusual camera noise, then Pfinder proceeds to analyze the region in more detail, and begins to build up a blob model of the person.