Defining and Refining Frameless Rendering

Abstract

Frame1ess Rendering (FR) is a rendering paradigm which performs stochastic temporal filtering by updating pixels in a random order, based on most recent available input data, and displaying them to the screen immediately (3]. It has inherent in its design, visual artifacts that come across as "spatial scatter''. Due to the asynchronicity of pixel updates, a noisy or "scattered" image is perceived if the update rate is not fast enough. We describe the fac1ors that influence the perceptual quality when Frarneless Rendering is implemented such as object and viewpoint velocities, pattern and object size, and pattern and object separation. The factors are formalized mathematically with respect to their impact on spatial scatter. Spatial scatter is defined formally as a combination of velocity, density and gradient. We also attempt to define Frameless Rendering in terms of a new graphics pipeline, and more concretely in terms of strategies for pixel prioritization and handling aging pixels. We zoom in on three factors that influence the perceptual quality. We compare scenes \with strong versus \weak temporal coherence, high versus low frequency content, and high versus low pixel computation rates. If strong temporal coherence is maintained, a Frameless Rendering sequence exhibits little image degradation. Cutaway scenes have no coherence between two successive frames and offer a natural dissolve. Weak temporal coherence is associated with decreased image fidelity. Image sequences comprised of scenes with primarily low frequency content exhibit very little image degradation when compared to image sequences comprised of scenes with high frequencies. Computation speeds of 30 M pixels/second, sampling inputs at JOO Hz, give the smoothness of motion and visual fidelity of a traditional 30 fps double-buffered animation. (See Section 7 for an anomalous case.) Compute speeds of I to 4 M pixels/second reduce latency and smooth motion seen in a comparable double buffered animation. We also show comparisons of Frameless Rendering with reduced pixel resolution. In cases of strong temporal coherence, Frameless Rendering exhibits relatively little image degradation in comparison with the reduced pixel resolution alternative. On the other hand, in cases of relatively weak temporal coherence, the lo\V resolution image sequence exhibits relatively little image degradation when compared with the Frameless Rendering.