A Shaped Temporal Filter Camera

In the time domain, movie cameras typically perform a discrete sampling of real-world imagery. While the effect of discrete sampling in the spatial domain is well understood, the effect on temporal sampling during acquisition has been rarely addressed in the past, partially due to the lack of configurable hardware. In this article, we investigate the effect of different temporal sampling kernels as prefilters and show that extended, overlapping kernels can mitigate aliasing artifacts. Further, NPR effects, such as enhanced motion blur, can be achieved. In addition, online Fourier transforms in the temporal domain provide a novel tool for analyzing and visualizing time dependent effects. An integrated system, consisting of a 500 Hz high speed camera and a desktop computer with an off-theshelf GPU, performs the necessary filtering or analysis in real-time and outputs the filtered result at video rates. We demonstrate the effect of different sampling kernels in creating enhanced movies and stills of fast motion.