THEME 2 OPERATION4

Operation 4: Visualization and rendering

Head : Jean-Michel Dischler PR1

Participants : Rémi Allègre AP, Basile Sauvage AP, Murielle Torregrossa associated AP, Olivier Génevaux, RI CNRS

Ph.D. candidates : Lucas Ammann, Alexandre Ancel

Postdocs : Guillaume Gilet, Frédéric Larue

The main goal of this operation consists in developing processing and rendering algorithms efficient for geometric entities issued from measures, by trying to : (1) develop simple procedures for the acquisition, particularly for the appearance of objects, (2) to enhance the visual quality of rendering in computer-generated images, and (3) enhance the exploration and visual manipulation of complex geometric entities, in particular volume entities.

Acquisition, analysis, synthesis and representation of objects appearance

In this scope we denote by appearance essentially the notion of texture [1-DGM07]. It consists of the feature patterns that we see appear on surfaces allowing to identify the material it is constituted with (wood, granite, organic tissue, etc.). We offer analysis and synthesis methods to reproduce colored patterns on surfaces starting from sample images [2-DZ06]. The project RIAM AMI3d in collaboration with Holo3 (regional center of technology transfer) aimed at acquiring work of art objects by including 4D (directional) texture information on top of geometric information. In this context we have developed an automatic process which doesn't need human interaction anymore for registering the images with the geometry [4-LD06]. The current project ANR ATROCO pursues this move on one side, by simplifying the process of acquisition (in particular by using digital images taken by standard devices) and on the other side by offering efficient techniques for handling huge data volumes coming from digitization (by using multi-scale representations or alternative ones well-suited for GPU rendering). The whole of this activity is enrolled in the digitization platform of IGG which has several measurement devices of which one laser scanner from LeicaGeosystems. Textures are also used to reproduce certain optical phenomena such as the refraction. We have proposed a new method based on spherical harmonics compression to reproduce this phenomenon in real-time [4-GLD06].

Direct volume visualization

The goal of this action, in relation with the program Réva and jointly with the research group ICPS at LSIIT (thesis of S. Marchesin co-advised with C. Mongenet), consists of increasing the interactivity of large 3D data visualization by direct volume rendering [4-MMD06, 4-MDM04, 4-MMD08] by exploiting GPU functionalities (with or without cluster) such as their capability to process 3D textures. The data comes from medical MRI, simulation codes developed by mathematicians and physicists from the project INRIA CALVI. This topic is structured around a ANR project of type massive data, MASSIM, whose goal was to conceive and realize a platform allowing to couple parallel and distributed simulations with a monitoring and interactive visualization system. The goal of this action also consists in enhancing the expressivity of the rendering of this data by using automatic mechanisms of transparency correction [4-MDM07]. This action goes on in the context of Alexandre Ancel's thesis (co-advising with Catherine Mongenet).