Static computer vision techniques enable non-intrusive observation and analysis of biometrics such as eye blinks. However, ambiguous eye behaviors such as partial blinks and asymmetric eyelid movements present problems that computer vision techniques relying on static appearance alone cannot solve reliably. Image flow analysis enables reliable and efficient interpretation of these ambiguous eye blink behaviors. In this paper we present a method for using image flow analysis to compute problematic eye blink parameters. The flow analysis produces the magnitude and direction of the eyelid movement. A deterministic finite state machine uses the eyelid movement data to compute blink parameters (e.g., blink count, blink rate, and other transitional statistics) for use in human computer interaction applications across a wide range of disciplines. We conducted extensive experiments employing this method on approximately 750K color video frames of five subjects.