Novel Temporal Views of Moving Objects for Gait Biometrics

There is increasing interest in novel view reconstruction but less for new time-based views of moving objects as needed for gait biometric deployment. Our interests concern reconstructing moving shapes from their moment history with a view to constructing new temporal views. This paper shows how the moment description through an object sequence can be used to predict missing or intermediate frames within the sequence. Additionally, this highlights generic aspects of moment reconstruction which rarely receive more than scant attention. We use Zernike moments for the convenience of reconstruction, although the framework is applicable to all types of moments. As an example, we show that by interpolating the moment history of a moving human silhouette, a missing frame can be constructed with accuracy, providing a practical basis for the construction of new temporal views of moving objects.