Formulas for description of center drift and for computing other geometric quantities of a binary object are derived for the purpose of estimating the location of the original object center, area of the object, and noise level in a noisy environment. Simulations support the theory. The coordinates of the object center estimated by the formulas are more accurate than those produced by median filtering. Consequently, more accurate object features are obtained. The formulas also lead us to conclude that when the object center coincides with the geometric center of the input plane, the object center does not drift, regardless of noise level. An optimal formation of concentric circles used for extraction of object features and for subpixel registration is also proposed.