andrew@dtg.nsc.com (Lord Snooty @ The Giant Poisoned Electric Head ) (07/15/90)
I would like to know if anyone could point me to NN methods used to solve the motion estimation problem. This problem concerns itself with finding the best estimate of displacement (expressed as a 2D vector) of non- overlapping blocks of pixels, whereby each block is surrounded by a (of course) larger "search window". I am particularly interested in approaches which exploit scale, translational and rotational invariance properties (_vide_ Widrow, Fukushima) to associatively identify the new blocks and thus to make the translational motion (or zoom or rotation) identifiable. Please e-mail responses. If there is interest, I will post a summary to the net. -- ........................................................................... Andrew Palfreyman "Smell isn't spatial or temporal. andrew@dtg.nsc.com It's like pure thought." - Gary Lynch