Motion estimation based on the energy flow

By taking pixels of image sequences as moving gas molecules, a novel concept, image temperature, is proposed to de-scribe the natural property of the image motion. The idea comes from the revelation of the Maxwell-Boltzmann Distribution Law in gas dy-namic theories. Furthermore, another concept of energy flow corresponding to the optical flow is developed, and the method of the energyflow equation (EFE) is established to estimate image motion. The experiment indicates a better performance of the proposed EFE schemewith significantly reduced false motion estimates when compared to the traditional optical flow equation (OFE).