Generating antialiased images at low sampling densities

Abstract

Ray tracing produces point samples of an image from a 3-D model. Constructing an antialiased digital picture from point samples is difficult without resorting to extremely high sampling densities. This paper describes a program that focuses on that problem. While it is impossible to totally eliminate aliasing, it has been shown that nonuniform sampling yields aliasing that is less conspicuous to the observer. An algorithm is presented for fast generation of nonuniform sampling patterns that are optimal in some sense. Some regions of an image may require extra sampling to avoid strong aliasing. Deciding where to do extra sampling can be guided by knowledge of how the eye perceives noise as a function of contrast and color. Finally, to generate the digital picture, the image must be reconstructed from the samples and resampled at the display pixel rate. The nonuniformity of the samples complicates this process, and a new nonuniform reconstruction filter is presented which solves this problem efficiently. This paper was presented in SIGGRAPH 87.

Recent Advances in Adaptive Sampling and Reconstruction for Monte Carlo Rendering

“…They perform wavelet analysis of the rendered image and analyze its error using Mitchell's contrast metric [Mit87]. They perform wavelet shrinkage [DJ94], that is, they subtract a noise estimate from individual wavelet coefficients, to minimize the error metric.…”

“…For example, Mitchell [Mit87] proposed a two-step approach to adaptively sample the image plane considering a contrast metric inspired by human perception, and he developed a reconstruction filter to deal with the nonuniform sample distributions. Parker and Sloan [PS89] and Guo [Guo98] sample the image plane using progressive refinement, and both apply polynomial reconstruction filters.…”

“…Hence, we call such methods "a posteriori". While this strategy goes back to the earliest adaptive sampling techniques [Mit87], Overbeck's approach combines it with more powerful reconstruction filters, which was a crucial step paving the way for further advances.…”

KD-tree based parallel adaptive rendering

“…Most researchers prefer to the first one which only adaptively samples the image plane. Early methods [1,6,7] generate adaptive samples relying on the measurement of local variance or attempt to use an density map of the image [8].…”

“…The adaptive sampling methods give better quality results than the images obtained by using stochastic sampling methods [1]. However, most of these methods only adaptively sample the image plane dimensions, but randomly sample the other dimensions.…”