Dynamic scene analysis from the polar form of image flow representations

A basic area of research in digital video processing is the frame to frame low level motion estimation of image sequences. The motion information obtained is usually displayed as a 2-D distribution of displacement vectors, called the optical flow field or image flow. How to use that information, in order to precisely estimate the location of objects in the scene, remains a relevant and very hard problem in computer vision. In previous schemes, the discontinuities appearing in the motion field have been studied in order to use them to discriminate between the various objects in the scene. In this work, the polar form of the image flow representation is used to enable a new approach to the handling of motion field discontinuities in dynamic scene analysis