Our approach is entirely image base

Our approach is entirely image based. Animators can adjust the extent of the generated blur to their liking after the images are acquired We take advantage of well-established computer vision techniques to segment and track large motions throughout a sequence. Our proposed techniques rely on combinations of back- ground subtraction and template matching in addition to the previously explored capabilities of optical flow and visual motion computation. Though existing techniques for encoding motions are becoming quite refined,such methods are not sufficient for our needs.