cv.jit.track Tracks the position of a number of pixels over time. Tracks the position of a number of pixels over time. This object uses the pyramidal implementation of the Lucas-Kanade technique described by Jean-Yves Bouguet. The Lucas-Kanade method of calculating optical flow, as stated above, is very precise for small movements but is incapable of accurately guessing large motions. The algorithm used here circumvents this problem by first guessing the position of the target pixel on a very low-resolution copy of the image (1/8 the original size in this implementation.) Even large movements will have only small pixel displacement on such a small matrix. This first guess is then used to make another guess on a larger copy, and the process is repeated until the new position is guessed on the original image. This technique is too computationally intensive to be used for every pixel, as in the case of cv.jit.Lkflow. However, it is perfect for a small set of target pixels.

from the library : cv.jit Format : External Environment : jitter / max / msp freeware Max 4.x http://www.iamas.ac.jp/~jovan02/cv/download.html