Efficient rendering of light and camera animation for navigating a frame array Alex M´ endez-Feliu Universitat de Girona amendez@ima.udg.es http://ima.udg.es/˜amendez/ Mateu Sbert Universitat de Girona mateu@ima.udg.es http://ima.udg.es/˜mateu/ Abstract Computing a series of frames of an animation in a global illumination framework is a very costly process. If we want in addition to experiment with different camera and lighting positions, the cost can become prohibitive. In this paper we show how this computation time can be dramat- ically reduced by combining different simplifi- cations and the reuse of previously computed data. On the one hand, the data between neigh- bor frames in a walkthrough (camera animation) can be reused by reprojection. On the other hand, using the obscurances technique, we sim- plify the computation of the global illumination by separating the direct and indirect lighting and by darkening the more occluded parts of the ob- jects of the scene. Color bleeding is also applied. Due to the separation of direct and indirect il- lumination, for a given camera position, the in- direct illumination needs to be computed only once, while the direct lighting can be computed for as many different light positions as we like. Both camera and light animations can be effi- ciently combined giving as result a set of images through which we can navigate and, for exam- ple, see the light animation for different camera positions, do the same walkthrough for different locations of the light, or combine both light and camera animations in the same movie. Keywords: Computer Graphics, Three- Dimensional Graphics and realism, Raytracing, Rendering, Animation 1 Introduction In production rendering of most short and long animated films and virtual scenarios of a movie the maximum realism has to be achieved. Nowa- days, the most recent global illumination tech- niques do the work, though they require a big ef- fort in computation power and time. Global illu- mination aims to simulate the lighting of a scene in a physically realistic way using ray casting based techniques, as path tracing [1] and pho- ton mapping [2]. In each frame millions of rays have to be cast to allow for an image without disturbing noise and without noticeable flicker- ing between frames. In this paper we will concentrate on two kinds of acceleration for global illumination: the reuse of information between contiguous frames in a walkthrough (camera animation) and the use of the obscurances technique, actually not a global illumination technique, but a powerful simpli- fication to compute indirect illumination that leads to very good visual results and allows to compute very efficiently the animation of light. Havran et al. in [3] presented an architecture for reusing information between frames based on reprojecting to neighbor eye positions the hit of the scene given by a ray cast from the eye through one pixel. In [4], Mendez et al. demon- strate that Havran et al.’s technique is biased for general BRDF’s and propose an unbiased solu- tion. They also showed that for diffuse BRDF’s, both theirs and Havran et al.’s solution are the same. We use this last one, an unweighted av-