Improving adaptive display with temporally adaptive rendering.

Abstract

Making computer imagery more responsive and realistic is one of the most basic goals of graphics researchers, and adaptive display is one of the primary means for achieving it. While previous displays have achieved a spatial adaptivity, our research focuses on achieving temporal adaptivity--sampling some regions not only more densely, but also more often. We use closed loop feedback to guide sampling to image regions that change significantly over space or time. Adaptive reconstruction emphasizes older samples in static settings, resulting in sharper images; and new samples in dynamic settings, resulting in images that may be blurry but are up-to-date. In terms of peak signal-to-noise ratio, this prototype produces much better image streams than nonadaptive renderers with the same simulated sampling rates. This new display also offers new opportunities for adapting to user state, allowing adaptive response both where and when it is needed. Our prototype system already responds interactively to changes in the user's viewpoint, it might also respond to any of a number of other indications of user state, including eye tracking, repeatedly manipulated objects, and biometrics.