This paper deals with the estimation of the time-to-contact in dynamic vision. It is well known that differential invariants of the image velocity field can be used to characterize the shape changes of objects in the scene, due to relative motion between the observer and the scene. Under the hypothesis of constant velocity along the optical axis, the time-to-contact turns out to be a function of the area enclosed by the object contour and its time derivative. In the paper, a novel approach based on set membership estimation theory is proposed to estimate the variables involved in the computation of the time-to-contact. Both errors in the motion model and image measurement noise are described as unknown-but-bounded disturbances, without requiring any statistical assumption. The proposed technique allows for the computation of guaranteed bounds on the time-to-contact estimates in finite time, a crucial issue in all problems where a robust evaluation of the time-to-contact is in order. (C) 2003 Elsevier Science Ltd. All rights reserved.