Jean-Michel DISCHLER

Jean-Michel DISCHLER
Professor at the University of Strasbourg.
E-mail: dischler -at- unistra -dot- fr
LSIIT, UMR CNRS-UDS 7005 (Laboratoire des Sciences de l'Image, de l'Informatique et de la Télédétection)
Pôle API, boulevard Sébastien Brant, F67400 ILLKIRCH Cedex,
Tel: (+33) 03 68 85 45 59 , Fax: (+33) 03 68 85 44 55

Full research-staff member of the laboratory LSIIT, and employed as professor at the department of computer science of the University of Strasbourg.

J.M. Dischler served as vice-director of the LSIIT Lab (from 2005 to 2008), managed the master degree in Image and Computing (IICI) from 2005 to 2009, and served as a member of the scientific council of the University Louis Pasteur before it became the University of Strasbourg. From September 2002 to August 2005, he was leading the department of Computer Science at the Faculty for Mathematics and Computer Science.
J.M. Dischler belongs to the 3D Computer Graphics Group, IGG, where he is supervizing the rendering and visualization research activities. He was an invited member of the INRIA Lorraine project CALVI (CALVI for CALcul scientifique et VIsualisation - Scientific Computation and Visualization) from 2003 to 2007 were he worked more specifically in the field of volume rendering. He left this group in 2007 to start activities in the field of 3D model / texture acquisition.
He completed a PhD in computer science at the University of Strasbourg (that he defended in January 1996). In January 2000, he obtained the french research supervision diploma: Habilitation a diriger des recherches, from the University of Limoges, where he remained an assistant professor during five years (1996-2001).

link to Publications

Teaching

data structures and advanced algorithmics, computer graphics, visualization, object oriented programming, history of computer science.

Research

(1) texture modeling and acquisition, texture analysis and synthesis;
(2) 3D acquisition for cultural heritage, processing of scanned data;
(3) interactive display using GPUs;
(4) direct volume rendering.

Some past and present research projects

	Pre-Integrated Volume Rendering with Non-Linear Gradient Interpolation, IEEE Vis 2010 (VisWeek 2010) Amel Guetat, Alexandre Ancel, Stephane Marchesin, and Jean-Michel Dischler Abstract. Shading is an important feature for the comprehension of volume datasets, but is difficult to implement accurately. Current techniques based on pre-integrated direct volume rendering approximate the volume rendering integral by ignoring non-linear gradient variations between front and back samples, which might result in cumulated shading errors when gradient variations are important and / or when the illumination function features high frequencies. In this paper, we explore a simple approach for pre-integrated volume rendering with non-linear gradient interpolation between front and back samples. We consider that the gradient smoothly varies along a quadratic curve instead of a segment in-between consecutive samples. This not only allows us to compute more accurate shaded pre-integrated look-up tables, but also allows us to more efficiently process shading amplifying effects, based on gradient filtering. An interesting property is that the pre-integration tables we use remain two-dimensional as for usual pre-integrated classification. We conduct experiments using a full hardware approach with the Blinn-Phong illumination model as well as with a non-photorealistic illumination model.
	An Image-Based Approach for Stochastic Volumetric and Procedural Details, CGF Vol.29(4), EGSR 2010 Guillaume Gilet, Jean-Michel Dischler Abstract. Noisy volumetric details like clouds, grounds, plaster, bark, roughcast, etc. are frequently encountered in nature and bring an important contribution to the realism of outdoor scenes. We introduce a new interactive approach, easing the creation of procedural representations of “stochastic” volumetric details by using a single example photograph. Instead of attempting to reconstruct an accurate geometric representation from the photograph, we use a stochastic multi-scale approach that fits parameters of a multi-layered noise-based 3D deformation model, using a multi-resolution filter banks error metric. Once computed, visually similar details can be applied to arbitrary objects with a high degree of visual realism, since lighting and parallax effects are naturally taken into account. Our approach is inspired by image-based techniques. In practice, the user supplies a photograph of an object covered by noisy details, provides a corresponding coarse approximation of the shape of this object as well as an estimated lighting condition (generally a light source direction). Our system then determines the corresponding noise-based representation as well as some diffuse, ambient, specular and semi-transparency reflectance parameters. The resulting details are fully procedural and, as such, have the advantage of extreme compactness, while they can be infinitely extended without repetition in order to cover huge surfaces.
	Feature-Driven Ambient Occlusion for Direct Volume Rendering, in EG/IEEE Volume Graphics 2010 Alexandre Ancel, Jean-Michel Dischler, Catherine Mongenet Abstract.Ambient occlusion techniques were introduced to improve data comprehension by bringing soft fading shadows. They consist in attenuating light by considering the occlusion resulting from the presence of neighboring structures. Recently introduced in volume rendering, we show that the straightforward application of ambient occlusion in direct volume rendering has its limits as rendering a multi-layer volume results in overdarkening the internal layers of the volume. This paper proposes to address the overdarkening issue by computing ambient occlusion according to the features present in the dataset. This allows us to neglect inter-occlusions between features without losing the auto-occlusions that give cues on the shape of the considered features. We use a GPU-based approach with bricking to speed up the computations of our ambient occlusion method. Results show that our approach not only improves the visual quality of images compared to classical ambient occlusion, but it is also less parametersensitive, thus furthermore improving usability for every-day users.
	Hybrid Rendering of Dynamic Heightfields using Ray-Casting and Mesh Rasterization, in GI 2010 Lucas Ammann, Olivier Génevaux, Jean-Michel Dischler Abstract.This paper presents a flexible hybrid method designed to render heightfield data, such as terrains, on GPU. It combines two traditional techniques, namely mesh-based rendering and per-pixel raycasting. A heuristic is proposed to dynamically choose between these two techniques. To balance rendering performance against quality, an adaptive mechanism is introduced that depends on viewing conditions and heightfield characteristics. It manages the precision of the ray-casting rendering, while mesh rendering is reserved for the finest level of details. Our method is GPU accelerated and achieves real-time rendering performance with high accuracy. Moreover, contrary to most terrains rendering methods, our technique does not rely on time-consuming pre-processing steps to update complex data structures. As a consequence, it gracefully handles dynamic heightfields, making it useful for interactive terrain edition or real-time simulation processes.
	Per-pixel Opacity Modulation for Feature Enhancement in Volume Rendering, IEEE TVCG Vol.16(4), 2010 Stephane Marchesin, Jean-Michel Dischler, Catherine Mongenet Abstract.Classical direct volume rendering techniques accumulate color and opacity contributions using the standard volume rendering equation approximated by alpha blending. However, such standard rendering techniques, often also aiming at visual realism, are not always adequate for efficient data exploration, especially when large opaque areas are present in a dataset, since such areas can occlude important features and make them invisible. On the other hand, the use of highly transparent transfer functions allows viewing all the features at once, but often makes these features barely visible. In order to enhance feature visibility, we present in this paper a straightforward rendering technique that consists in modifying the traditional volume rendering equation independently of any transfer function. Our approach is fully automatic and based on a function quantifying the relative importance of each voxel in the final rendering called relevance function. This function is subsequently used to dynamically adjust the opacity of the contributions per-pixel. As will be shown by our comparative study with standard volume rendering, this makes our rendering method much more suitable for interactive data exploration at a low extra cost. Thereby, our method avoids feature visibility restrictions without relying on a transfer function and yet maintains a visual similarity with standard volume rendering.
	A Framework for Interactive Hypertexture Modelling, CGF Vol. 28(8), 2009 Guillaume Gilet, Jean-Michel Dischler Abstract. Hypertexturing can be a powerful way of adding rich geometric details to surfaces at low memory cost by using a procedural 3D space distortion. However, this special kind of texturing technique still raises a major problem: the efficient control of the visual result. In this paper, we introduce a framework for interactive hypertexture modeling. This framework is based on two contributions. Firstly, we propose a reformulation of the density modulation function. Our density modulation is based on the notion of shape transfer function. This function, which can be easily edited by users, allows us to control in an intuitive way the visual appearance of the geometric details resulting from the space distortion. Secondly, we propose to use a hybrid surface and volume-point-based representation in order to be able to dynamically hypertexture arbitrary objects at interactive frame rates. The rendering consists in a combined splat- and raycasting-based direct volume rendering technique. The splats are used to model the volumetric object while raycasting allows us to add the details. An experimental study on users shows that our approach improves the design of hypertextures and yet preserves their procedural nature.
	Second Order Pre-Integrated Volume Rendering, IEEE VGTC Pacific Visualization Symposium 2008, PacificVis 2008 Jean-François El Hajjar, Stephane Marchesin, Jean-Michel Dischler, Catherine Mongenet Abstract. In the field of Volume Rendering, pre-integration techniques for arbitrary transfer functions has certainly led to the most significant and convincing results both quality and performance wise on standard PC consumer graphics. By showing that the ideal scalar signal along the cast rays is better approximated by a succession of polynomial curves as opposed to linear segments, we propose a new method for pre-integrated volume rendering. This method is based on a second order polynomial interpolation of the scalar values, allowing it to converge more rapidly towards the integration of a volume reconstructed by a trilinear filter. This approach manages to capture the smoothness of the volume’s details without the need of further ray resampling, and consequently succeeds in reducing the visual artefacts in comparison to previous techniques. Futhermore, we adapt an existing technique to compute our pre-integration tables using the GPU, thus making our approach suitable for transfer function manipulations.
	Feature Enhancement using Locally Adaptive Volume Rendering, IEEE/EG Volume Graphics, 2007 Stéphane Marchesin and Jean-Michel Dischler and Catherine Mongenet Abstract. Classical direct volume rendering techniques accumulate color and opacity contributions using the standard volume rendering equation approximated by alpha blending. However, such standard rendering techniques, often also aiming at visual realism, are not always adequate for efficient data exploration, especially when large opaque areas are present in a dataset, since such areas can occlude important features and make them invisible. On the other hand, the use of highly transparent transfer functions allows viewing all the features at once, but often makes these features barely visible. In this paper we introduce a new, straightforward rendering technique called locally adaptive volume rendering, that consists in slightly modifying the traditional volume rendering equation in order to improve the visibility of the features, independently of any transfer function. Our approach is fully automatic and based only on an initial binary classification of empty areas. This classification is used to dynamically adjust the opacity of the contributions per-pixel depending on the number of non-empty contributions to that pixel. As will be shown by our comparative study with standard volume rendering, this makes our rendering method much more suitable for interactive data exploration at a low extra cost. Thereby, our method avoids feature visibility restrictions without relying on a transfer function and yet maintains a visual similarity with standard volume rendering.
	A Pipeline for the Digitisation and the Realistic Rendering of Paintings, VAST 2007 F. Larue, L. Ammann, J.-M. Dischler Abstract. Digitization and visualization are both of great importance for Cultural Heritage, for instance for the design of virtual galleries. Despite a lot of research, enabling a real-time walkthrough around complex digital copies still remains difficult and challenging in the general case due to the complexity of the measurement and to the amount of data that has to be dealt with. In this paper, we introduce a new dedicated pipeline for both digitization and realistic rendering of art paintings. We exploit the fact that geometrical variations over the canvas are generally small, yet not negligible from a visual point of view. Unlike most existing painting digitization systems, we thus propose to acquire both geometry and texture. Then, we render both as a whole by using, for the texture, an analytical model which is fitted from real measurements, and by using for the geometry a hybrid approach combining two relief rendering techniques according to the scale. This allows us to derive an efficient adaptive scheme guaranteeing fast rendering rates for all viewpoints. With our pipeline the painting’s relief is well preserved, thus the rendering is of high quality, and in addition the final data representing the digital copies remain compact.
	A new vector field distance transform and its application to mesh processing from 3D scanned data, The Visual Computer, 2007 Marc Fournier · Jean-Michel Dischler · Dominique Bechmann Abstract We define a new 3D vector field distance transform to implicitly represent a mesh surface. We show this new representation is more accurate than the classic scalar field distance transform by comparing both representations with an error metric evaluation. Widely used Marching Cube triangulation algorithm is adapted to the new vector field distance transform to correctly reconstruct the resulting explicit surface. In the reconstruction process of 3D scanned data, useful mesh denoising operation is extended to the new vector field representation which enables adaptive and selective filtering features. Results show mesh processing with this new vector field representation is more accurate than with the scalar field distance transform and it outperforms previous mesh filtering algorithms. Future works are discussed to extend this new vector field representation to other mesh useful operations and applications.
	Automatic Registration and Calibration for Efficient Surface Light Field Acquisition, in VAST 2006 Frédéric Larue, Jean-Michel Dischler Abstract We present a novel protocol for the acquisition of surface light fields which is designed to deal with delicate objects that might not be touched or moved. This constraint is particularly important when art pieces are involved. Our protocol enables the automatic reconstruction of a model from many range images and the automatic registration of many pictures with the acquired geometry. A structured light pattern is first used to project a parameterization over the analyzed surface. Each surface point hit by this parameterization is uniquely identified, independently of the chosen viewpoint, and the problem of finding point-point and point-pixel correspondences is then immediately solved. These correspondences are finally used to perform the registrations and camera calibrations that provide the data to be used by a surface light field renderer. Video (AVI)
	Dynamic Load Balancing for Parallel Volume Rendering, in EGPGV 2006 Stéphane Marchesin and Catherine Mongenet and Jean-Michel Dischler Abstract Parallel volume rendering is one of the most efficient techniques to achieve real time visualization of large datasets by distributing the data and the rendering process over a cluster of machines. However, when using level of detail techniques or when zooming on parts of the datasets, load unbalance becomes a challenging issue that has not been widely studied in the context of hardware-based rendering. In this paper, we address this issue and show how to achieve good load balancing for parallel level of detail volume rendering. We do so by dynamically distributing the data among the rendering nodes according to the load of the previous frame. We illustrate the efficiency of our technique on large datasets.
	Real-Time Structured Texture Synthesis and Editing Using Image - Mesh Analogies, The Visual Computer 2006 Jean-Michel Dischler · Florence Zara Abstract We present a novel texture synthesis technique designed to reproduce at real-time frame-rates example texture images, with a special focus on patterns characterized by structural arrangements. Unlike current pixel-, patch- or texton-based schemes, that operate in image-space, our approach is structural. We propose to assimilate texture images to corresponding 2D geometric meshes (called texturemeshes). Our analysis mainly consists in generating automatically these meshes, while synthesis is then based on the creation of new vertex/polygon distributions matching some arrangement map. The output texture-image is obtained by rasterizing the previously generated polygons using graphics hardware capabilities, which guarantees high speed performance. By operating in geometry space instead of image / pixel-space, the proposed structural approach has a major advantage over current techniques: beyond pure texture reproduction, it permits us defining various tools, which allow users to further modify locally or globally and in realtime structural components of textures. By controlling the arrangement map, users can substitute new meshes in order to completely modify the structural appearance of input textures, yet maintaining a certain visual resemblance with the initial example image.
	Interactive Refraction on Complex Static Geometry using Spherical Harmonics, in I3DG 2006 Olivier Génevaux, Frédéric Larue, Jean-Michel Dischler Abstract Accurate refraction, thanks to raytracing, has always been a popular effect in computer graphics. However, devising a technique that produces realistic refractions at interactive rates remains an open problem. In this paper, a method to achieve realistic and interactive refractive effects through complex static geometry is proposed. It relies on an offline step where many light paths through the object are preevaluated. During rendering, these precomputed paths are used to provide approximations of actual refracted paths through the geometry, enabling further sampling of an environment map. The relevant information of the light paths, namely final output direction when leaving refractive object, is compressed using frequency domain based spherical harmonics. The matching decompression procedure, entirely offloaded onto graphics hardware, is handled at interactive speed. Video (AVI)
	Real-Time High-Quality View-Dependent Texture Mapping using Per-Pixel Visibility, in GRAPHITE 2005 Damien Porquet, Jean-Michel Dischler, Djamchid Ghazanfarpour Abstract We present an extension of View-Dependent Texture Mapping (VDTM) allowing rendering of complex geometric meshes at high frame rates without usual blurring or skinning artifacts. We combine a hybrid geometric and image-based representation of a given 3D object to speed-up rendering at the cost of a little loss of visual accuracy. During a precomputation step, we store an image-based version of the original mesh by simply and quickly computing textures from viewpoints positionned around it by the user. During the rendering step, we use these textures in order to map on the fly colors and geometric details onto the surface of a low-polygon-count version of the mesh. Real-time rendering is achieved while combining up to three viewpoints at a time, using pixel shaders. No parameterization of the mesh is needed and occlusion effects are taken into account while computing on the fly the best viewpoints for a given pixel. Moreover, the integration of this method in common real-time rendering systems is straightforward and allows applying self-shadowing as well as other z-buffer effects. Video (AVI)
	3D ROAM for Scalable Volume Visualization, in IEEE Symposium on Volume Visualization 2004 Stéphane Marchesin, Jean-Michel Dischler, Catherine Mongenet Abstract The 2D real time optimally adapting meshes (ROAM) algorithm has had wide success in the eld of terrain visualization, because of its efcient error-controlling properties. In this paper, we propose a generalization of ROAM in 3D suitable for scalable volume visualization. Therefore, we perform a straightforward 2D/3D analogy, replacing the triangle of 2D ROAM by its 3D equivalent, the tetrahedron. Although work in the eld of hierarchical tetrahedral meshes was widely undertaken, the produced meshes were not used for volumetric rendering purposes. We explain how to compute a bounded error inside the tetrahedron to build a hierarchical tetrahedral mesh and how to rene this mesh in real time to adapt it to the viewing conditions. We further show how to achieve cell sorting in linear time, thus yielding real time view-dependent display of the volumetric object. We present examples of large volume data sets and compare our approach with a similar one. Our results outline the high quality and computational efciency of our approach.
	Efficient light transport using pre-computed visibility, in IEEE Computer Graphics and Applications 23(3), 2003 K. Daubert, W. Heidrich, J. Kautz, J-M. Dischler and H.P. Seidel Abstract Visibility computations are the most time-consuming part of global illumination algorithms. The cost is amplified by the fact that identical or similar information is often recomputed multiple times. In particular, this is the case when multiple images of the same scene need to be generated under varying lighting conditions and/or viewpoints.This article describes a general method of precomputing, storing, and reusing visibility information for light transport in a number of different types of scenes. In particular, it considers general parametric surfaces, triangle meshes without a global parameterization, and participating media. It also reorders the light transport in such a way that the visibility information is accessed in structured memory access patterns.This yields a method well suited for SIMD-style parallelization of the light transport, and can efficiently be implemented both in software and graphics hardware.
	Simulating Fluid-Solid Interaction, in GI 2003 Olivier Génevaux, Arash Habibi, Jean-Michel Dischler Abstract Though realistic eulerian fluid simulation systems now provide believable movements, straightforward renderable surface representation, and affordable computation costs, they are still unable to deal with non-static objects in a realistic manner. Namely, objects can not have an influence on the fluid and be simultaneously affected by the fluid’s motion. In this paper, a simulation scheme for fluids allowing automatic generation of physically plausible motions alongside realistic interactions with solids is proposed. The method relies mainly on the definition of a coupling force between the solids and the fluid, thus bridging the gap between commonly used eulerian fluid animation models and lagrangian solid ones. This new method thus improves existing fluid simulations, making them capable of generating new kinds of motions, such as a floating ball displaced by the wave created thanks to its own splash into the water. Video1, Video2, Video3
	Texture Particles, Computer Graphics Forum (Eurographics) 2002 Jean-Michel Dischler, Karl Maritaud, Bruno Lévy, Djamchid Ghazanfarpour Abstract We present an analytical extension of texture synthesis techniques based on the distribution of elementary texture components. Our approach is similar to the bombing, cellular, macrostructured and lapped textures techniques, but provides the user with more control on both the texture analysis and synthesis phases. Therefore, high quality results can be obtained for a large number of structured or stochastic textures (bricks, marble, lawn, etc.). The analysis consists in decomposing textures into elementary components – that we call “texture particles” – and for which we analyze their specific spatial arrangements. The synthesis then consists in recomposing similar textures directly on arbitrary surfaces by taking into account the previously computed arrangements, extended to 3D surfaces. Compared to “pixel-based” analysis and synthesis methods, which have been recently generalized to arbitrary surfaces, our approach has three major advantages: (1) it is fast, which allows the user to interactively control the synthesis process. This further allows us to propose a large number of tools, granting a high degree of artistic freedom to the user. (2) It avoids the visual deterioration of the texture components by preserving their shapes as well as their spatial arrangements. (3) The texture particles can be not only images, but also 3D geometric elements, which extends significantly the domain of application.
	Coherent bump map recovery from a single image, in GI 2002 Jean-Michel Dischler, Karl Maritaud, Djamchid Ghazanfarpour Abstract In order to texture surfaces realistically with texture images (e.g. photos), it is important to consider the underlying relief. Here, a method is proposed to recover a coherent bump map from a single texture image. Different visual zones are first identified using segmentation and classification. Then, by linearly separating the relief into a noise-like small-scale component and a smooth “shape-related” large-scale component, we can automatically deduce the bump map as well as an “un-shaded” color map of the texture. The major advantage of our approach, compared to sophisticated measurement techniques based on multiple photos or specific devices, is its practical simplicity and broad accessibility, while it allows us to obtain very easily, via basic bump mapping or displacement mapping, rendering results of good quality.

Last update september 2010