About

PrISE-3D for "Perception, Interactions et Simulation d’Eclairage 3D" is a French ANR project.

The production of photorealistic computer-generated images requires the use of lighting simulation methods, characterised by the presence of perceptible noise. This noise gradually disappears according to the convergence of the methods used, but at the cost of very high calculation time. This is all the more penalizing for the production of sequences of 3D images, which require the calculation of several distinct images. The main goal of the PrISE-3D project is a better understanding of the perceptual mechanisms underlying the perception of a quality image, when faced with various stereoscopic restitution devices. For this purpose, procedures for identifying the perceptual thresholds of visual convergence of lighting simulation algorithms will be studied in the case of images in relief. These thresholds will then make it possible to determine criteria for automatically stopping the calculations, which can be used in the following cases the framework for high quality audiovisual productions, but also to guide the rendering of interactive algorithms, according to the actions carried out by a user in large immersive environments.

Articles

Jan. 6, 2021

Stopping criterion during rendering of Computer-Generated Images based on SVD-Entropy

The estimation of image quality and noise perception still remains an important issue in various image processing applications. It also became a hot topic in the field of photo-realistic computer graphics where noise is inherent in the calculation process. Unlike natural-scene images however, a reference image is not available for computer-generated images. Thus, classical methods to assess noise quantity and stopping criterion during the rendering process are not usable.

Projects:

@prise-3D

Quick overview of all PrISE-3D projects.

[P3D] pbrt-v4

Custom version of the early release of pbrt-v4 for the PrISE-3D project, the rendering system fourth edition of Physically Based Rendering: From Theory to Implementation.

vpbrt

Scene viewer for pbrt-v3 which allows the extraction of camera information relative to the current point of view of the scene

Resources

SIN3D dataset

This dataset is composed of 80 points of view of photo realistics images with different level of samples for each. The most noisy image is of 20 samples and the reference one (the most converged image obtained) is of 10000 samples.