A perceptual stopping condition for global illumination computations

Abstract

The aim of realistic image synthesis is to produce high fidelity images that authentically represent real scenes. As these images are produced for human observers, we may exploit the fact that not everything is perceived when viewing scene with our eyes. Thus, it is clear that taking advantage of the limited capacity of the human visual system (HVS), can significantly contribute to optimize rendering software.Global illumination methods are used to simulate realistic lighting in 3D scenes. They generally provide a progressive convergence to high-quality solution. One of the problem of such algorithms is to determine a stopping condition, for deciding if calculations reached a satisfactory convergence allowing the process to terminate.In this paper, we propose and we discuss different solutions to this important problem. We show different techniques based on the Visual Difference Predictor (VDP) proposed by Daly [Daly 1993] to define a perceptual stopping condition for rendering computations. We use the VDP to measure the perceived differences between rendered images and to guide the Path Tracing rendering to satisfy a perceptual quality. Also, in a controlled experimental setting with real subjects, we validate our results.

Perception of noise and global illumination: Toward an automatic stopping criterion based on SVM

“…Myszkowski [3] used the VDP principle to provide quantitative measures of perceptual convergence by predicting and estimating the perceivable differences between intermediate and final images. A similar approach was proposed by Takouachet and co-authors [18], where the VDP is used for estimating the differences between the initial very noisy image and the successive images of the progressive rendering process. In both approaches, the VDP only operates on the luminance channel and is costly to computed.…”

“…Recently VDP-based approaches [3,18] have been applied for detecting perceptual noise in progressives GI methods. Even if the VDP is an HVS model, its application remains costly in terms of computation time.…”

“…Previous works attempted to discover how to efficiently stop the Path Tracing based methods. In [3] and [18], the authors suggested the VDP for identifying when noise is visually not perceptible. We compared the results of our approach with the ones found by these two approaches.…”

“…Furthermore, the suggested model has the advantage of working on sub-images allowing therefore, a better distribution of the computation load. Actually, the approaches proposed in [3] and [18] require the use of the VDP, which is generally more costly to compute than the questioning of the model performed in our approach.…”

Image noise detection in global illumination methods based on FRVM

“…They used the visible differences predictor (VDP) algorithm to provide the quantitative measures of perceptual convergence by predicting and estimating the perceivable differences between the intermediate and the reference images. A similar approach has been proposed by Takouachet et al [6]. The VDP was used for estimating the differences between the first very noisy image and the successive images of the progressive rendering process.…”