Classifying Moving Objects as Rigid or Non-rigid without Correspondences

This paper presents a method for classifying moving objects as rigid or non-rigid based on the similarity of their appearance over multiple frames. This classiication is performed by using an image of the object obtained earlier in the video sequence and examining the uc-tuation in the similarity between this reference image and subsequent images of the object. To do this, we use a representation similar to that used by others to characterize the periodicity of an object's appearance. The hypothesis is that the appearance of rigid objects, under viewing conditions similar to orthographic projection, changes much more slowly than that of most non-rigid animate objects. The resulting algorithm is amenable to real-time on-line implementation .