Rotational motion estimation and analysis in digital image sequences

In this paper we consider the problem of motion estimation, analysis and selective reconstruction of objects undergoing rotational motion. There are multiple objects rotating with different angular velocities in the image sequences. The goal is to estimate their distinct motion parameters(in this case is the angular velocities), and also identify their locations at each time instance by selective reconstruction. These parameters and locations can be used for various purpose such as motion compensation in video coding, military target detection, recognition and tracking, robot's focus/shift attention, road navigation, etc. The new algorithm we have developed is based on angular velocity tuned 2D+T filters. One of the important fact about our algorithm is that it is effective for both spinning motion and orbiting motion, thus unifies the treatment of the two kinds of rotational motion. Furthermore, it can also derives the angular velocities seperatly by choosing different scale values once the objects have different sizes. The algorithm is robust against noise and occlusion. It is simulated and tested on several synthesized image sequences corrupted by noise and shows to be accurate.