A new algorithm for speckle tracking is introduced. The algorithm is based on 2D correlation processing of beamformed radio frequency (RF) data from a medical imaging scanner employing a linear array of transducers. However, it is based on two powerful features for mask size design and for optimizing search direction for efficient implementation. An eigenvalue decomposition of a local intraframe spatial correlation matrix is used to determine the mask size. In addition, the search direction is detrmined from the 2D gradiant of the phase profiles of an interframe spatial correlation matrix. The result is a robust and computationally efficient speckle tracking algorithm suitable for medical imaging applications. A description of the new algorithm is given in this paper along with illustrative examples from image sequences obtained from tissue mimicking phantoms and in vivo liver images from a healthy volunteer.