OPTICAL FLOW ESTIMATION OF THE HEART MOTION USING LINE PROCESS

This paper uses line process technique from computer vi-sion to enhance optical flow computation for the problem of cardiac motion estimation. The basic idea is to intro-duce the line process as a tool for handling discontinuities of the optical flow field. Optical flow showing cardiac mo-tion can then become piecewise smooth instead of globally smooth. Points of interest usually lie on the boundaries of the heart and this method is especially accurate at such points. The general problem is stated as a Bayes estima-tor and uses MRF framework to encode a priori knowl-edge. The MAP estimation is found as the minimum of the non-convex energy function using Highest Confidence First (HCF) algorithm. The advantages of HCF algorithm are that it is deterministic and the result is not dependant on the initialization step. The procedure is applied to ECG-gated MR image sequence of the beating heart