We present a method for modeling and rendering irregular and heterogeneous glass objects, with a specific emphasis on stained glass windows and window works often encountered in architecture from middle age to 18th century. The artisanal production of sheet glass results in glass panels displaying a vast variety of surface and volume irregularities like bubbles, irregular surface or smoothly varying refractive index, all of which contribute to the specific visual aspect of old glass. We propose to account for all the aforementioned effects in a unified framework based on signed distance functions and an analytic solution of the ray tracing equations on tetrahedral volume elements. We demonstrate how to construct an unbiased estimator for the transmitted lighting produced by such panels by using Fermat's principle and results from seismic ray theory. We use texture coordinates to map arbitrary sections of a complex glass panel onto the individual faces of a mesh, allowing the modeling and rendering of complex 3-dimensional objects composed of colored glass facets such as stained glass windows.

Estimating the features to be extracted from an image for classification tasks are sometimes difficult, especially if images are related to a particular kind of noise. The aim of this paper is to propose a neural network architecture named Guided-Generative Network (GGN) to extract refined information that allows to correctly quantify the noise present in a sliding window of images. GNN tends to find the desired features to address such a problem in order to emit a detection criterion of this noise. The proposed GGN is applied on photorealistic images which are rendered by Monte-Carlo methods by evaluating a large number of samples per pixel. An insufficient number of samples per pixel tends to result in residual noise which is very noticeable to humans. This noise can be reduced by increasing the number of samples, as proven by Monte-Carlo theory, but this involves considerable computational time. Finding the right number of samples needed for human observers to perceive no noise is still an open problem. The results obtained show that GGN can correctly solve the problem without prior knowledge of the noise while being competitive with existing methods.

This paper introduces a new algorithm, namely the equi-correlation network (ECON), to perform supervised classification, and regression. ECON is a kernelized LARS-like algorithm, by which we mean that ECON uses an l1regularization to produce sparse estimators, ECON efficiently rides the regularization path to obtain the estimator associated to any regularization constant values, and ECON represents the data by way of features induced by a feature function. The originality of ECON is that it automatically tunes the parameters of the features while riding the regularization path. So, ECON has the unique ability to produce optimally tuned features for each value of the constant of regularization. We illustrate the remarkable experimental performance of ECON on standard benchmark datasets; we also present a novel application of machine learning in the field of computer graphics, namely the approximation of photometric solids.

Catopsys is a low-cost projection system aiming at making mixed reality (virtual, augmented or diminished reality) affordable. It combines a videoprojector, a camera and a convex mirror and works in a non-specific room. This system displays an immersive environment by projecting an image onto the different parts of the room. However, the presence of an uncalibrated projector, heterogeneous materials and light inter-reflections influence the colors of the environment displayed in the room. Radiometric compensation of the projection process enables the system to reduce this problem.

In this paper we study both theoretically and experimentally some properties of classical linear systems solvers, according to the radiosity assumptions. We prove important properties for some of these solvers which allow the user to choose the best one. We then introduce a new technique so called hybridization whose purpose is to increase the convergence speed of iterative methods. It provides very efficient results for the well-known Gauss-Seidel solver. This technique has been successfully applied to both a group progressive radiosity approach and a full-matrix radiosity method which has been specifically designed for plant growth simulation. 1.

The z-buffer is a well-known hidden-part removal technique commonly used by local illumination algorithms. Some global illumination approaches use this technique too, in order to approximate energy exchanges. In this paper we propose a massively parallel implementation of the z-buffer on the MP-1 machine. Efficiency is achieved by precisely studying the different stages of the algorithm, and by taking care in correctly using the SIMD control of the architecture. Local illumination models are then applied to the z-buffer algorithm, by using Gouraud and Phong's interpolations. Finally, the parallel z-buffer is used in a massively parallel radiosity algorithm. The results obtained allow to provide quickly illuminated images by decreasing dramatically the computation time required for global illumination.