In this paper we present a novel method for estimation of blood flow speed and vessel location from thermal video. The method is based on a bioheat transfer model that reflects the thermo-physiological processes in a skin region proximal to a major vessel. The model assumes the form of a partial differential equation (PDE) with boundary conditions. Initially, we test the soundness of our model by performing direct numerical simulation. Then, we solve the inverse problem both in steady and dynamic states on data provided by a thermal imaging system. Our method opens exciting possibilities in biometrics and biomedicine. Among others, it promises to revolutionize polygraph examinations by eliminating wiring and improving accuracy. It also establishes the feasibility of continuous 2D physiological monitoring of human patients in a contact-free manner.