Linear perspective projections are used extensively in graphics. They provide a non-distorted view, with simple computations that map easily to hardware. Non-linear projections, such as the view given by a fish-eye lens are also used, either for artistic reasons or in order to provide a larger field of view, e.g. to approximate environment reflections or omnidirectional shadow maps. As the computations related to non-linear projections are more involved, they are harder to implement, especially in hardware, and have found little use so far in practical applications. In this paper, we apply existing methods for non-linear projections [Lloyd et al. 2006; Hou et al. 2006; Fournier 2005] to a specific class: non-linear projections with a single center of projection, radial symmetry and convexity. This class includes, but is not limited to, paraboloid projections, hemispherical projections and fish-eye lenses. We show that, for this class, the projection of a 3D triangle is a single curved triangle, and we give a mathematical analysis of the curved edges of the triangle; this analysis allows us to reduce the computations involved, and to provide a faster implementation. The overhead for non-linearity is bearable and can be balanced with the fact that a single nonlinear projection can replaces as many as five linear projections (in a hemicube), with less discontinuities and a smaller memory cost, thus making non-linear projections a practical alternative.

Mediated reality describes the concept of filtering or vision of reality, typically using a head-worn video mixing display. In this paper, we propose a generalized concept and new tools for interactively mediated reality. We present also our first prototype system for painting, grabbing and gluing together real and virtual elements.

Abstract It has been shown that an inaccurate motor planning of a fast goal‐directed movement could be amended through an online comparison between the goal and the instantaneous representation of the (unseen) moving hand. The aim of this study was to test whether such an optimum processing was still efficient in a virtual visual environment (VE), by exploring the effect of immersion on the online visuomotor control of goal‐directed arm movement. Six subjects participated in the experiment, in which hand reaching toward either a stationary or a moving virtual visual target (double‐step paradigm) was executed in an upright posture. The analysis of the hand kinematics and postural adjustments showed an accurate control of movement and balance in both reaching conditions. This indicates that immersion does not affect the online control processes used by the CNS to adjust movement in interactive VE. Copyright © 2003 John Wiley & Sons, Ltd.

The client-server architecture is widely used in the construction of virtual environments because it allows specific services to be provided to the connected clients and management tasks to be performed, such as reducing the network load by performing message filtering. The majority of client-server systems are incompatible with each other, both in terms of hardware and software. In particular, the different communication protocols make the cooperation between the various systems difficult. This paper guides the reader through the construction and implementation of a server that consists of clearly distinct components, developed to reflect a reference model for a generic and open client-server system. This server, employed in the PAVR network, not only allows incompatible clients and applications to be linked into a common environment with little effort, but, in addition, the detailed description of its implementation can provide valuable help to whoever undertakes the task of coding such systems.

The continuously increasing complexity of computer animations makes it necessary to rely on the knowledge of various experts to cover the different areas of computer graphics and animation. This fact leads to increasing effort being put into research concerning cooperative working over the Internet. However, it still requires substantial effort and time to combine different animation techniques in a common virtual environment. When trying to perform collaborative animation over a network, we often face the problem of having to combine animation systems and applications based on different software and hardware and using incompatible data structures. We present an approach, based on a client-server architecture and employing a VRML based language as common interchange format, that allows inhomogeneous systems to be easily incorporated into a collaborative animation. The applications can be freed from employing plug-ins or libraries to link into a common animation platform; they keep a local copy of the global scene and only need the ability to export the internal data representation into the so called "PaVRML" language, the language used to exchange data and synchronize clients. This approach allows a number of practitioners to share their know-how within a common animation without requiring the huge amount of work necessary to port their application to a common platform. It also makes it possible in the first place to combine the capabilities of different animation systems into a single complex animation. We also investigate solutions to optimize the network load for real time applications. We present preliminary results and discuss the future developments of this ongoing work.

This paper presents an integrated set of tools to describe movements of objects in a framework of key-framed animation, both in simple and complex scenes. We want to have a powerful, robust and fast tool for designing either quickly prototyped or finely tuned motions. To achieve that, we bring together Hermite splines (with an optional simplified interface for tangents) and logarithms of quaternions for orientations. We define an interface featuring both by-hand and simplified tangents edition. This gives us a unified simple interface for both translations and orientations. The construction of complex movements is easier to perform by assembling simpler ones, through hierarchy, rendezvous or links with other types of animation (dynamic, articulated structures, ...). We particularly address implementation considerations, and the problems of fast computations, to get interactivity.>

Abstract This paper describes a method that facilitates the use of physically‐based models by animators. The main point is to give the animator a familiar interface, while providing a simulation module which detects collisions, thus enhancing realism. The user gives a set of key‐frames to guide motion, but does not have to address problems such as interpenetration avoidance, deformations due to collisions, or realism of motion. The simulator will correct the trajectories and compute deformations according to each object's physical properties (such as mass, inertia, stiffness) as well as the collisions and contacts automatically detected during motion. To achieve this, objects are provided with actuators capable of generating forces and torques computed via generalized proportional‐derivative controllers. When deflected by external actions, actuated objects try to return to their initial path. Speed variations over time are computed during the simulation, and depend on the complexity of the paths, on the objects models, and on the events such as collisions occurring during motion. In addition simulations are generated at interactive rates, even in the case of complex articulated objects. This facilitates the fine tuning of an animation sequence.

This paper presents a built-in mechanism for automatic detection of prototypes (as opposed to parasite attractors) in a Hopfield network. It has a good statistical performance and avoids host computer overhead for this purpose. This mechanism is based on an internal coding of the prototypes during learning, using cyclic redundancy codes, and leads to an efficient implementation in VLSI. The immediate separation of prototypes from parasite attractors can be used to enhance the autonomy of the networks, aiming at hierarchical multinetwork structures. As an example, the use of such an architecture for the classification of handwritten digits is described.

The definition of the requirements for the design of a neural network associative memory, with on-chip training, in standard digital CMOS technology is addressed. Various learning rules that can be integrated in silicon and the associative memory properties of the resulting networks are investigated. The relationships between the architecture of the circuit and the learning rule are studied in order to minimize the extra circuitry required for the implementation of training. A 64-neuron associative memory with on-chip training has been manufactured, and its future extensions are outlined. Beyond the application to the specific circuit described, the general methodology for determining the accuracy requirements can be applied to other circuits and to other autoassociative memory architectures.