A New Way to Re-using Paths Qing Xu 1 and Mateu Sbert 2 1 Tianjin University, Tianjin 300072, China qingxu@tju.edu.cn 2 University of Girona, Girona 17003, Spain mateu@ima.udg.es Abstract. Monte Carlo is the only choice of physically correct method to compute the problem of global illumination in the field of realistic im- age synthesis. Reusing light transport paths is an interesting and effective tool to eliminate noise, which is one of the main problems of Monte Carlo based global illumination algorithms, such as Monte Carlo ray tracing. But reusing paths technique tends to group spike noise to form noise patches in the images. We propose an alternative way to implementing the reuse of paths to tackle this problem in this paper. Experimental results show that our new way is very promising. Keywords: Reusing paths, Monte Carlo, Global Illumination, Ray tracing. 1 Introduction Global illumination plays an important role in realistic image synthesis, and Monte Carlo based algorithms are the unique choice of physically correct meth- ods to compute the problem of global illumination [2]. The general Monte Carlo global illumination methods, including both the view dependent solutions and the final gathering schemes involved in Photon Mapping [3], usually employ the baseline Monte Carlo ray tracing (MCRT) [4] to produce the synthetic images pixel by pixel. MCRT uses sample paths through a pixel to calculate the pixel value by averaging the sample values, namely the light transport contributions of sample paths. Generally, MCRT gives very noisy images when inadequate sam- ples are used because of the slow convergence of the Monte Carlo techniques. Taking a large enough number of samples for each pixel can render images with- out noise, but it is too time consuming. A lot of progress has been made on reducing noise for images produced by Monte Carlo based global illumination algorithms. Bidirectional path tracing [5] [6] and Metropolis light transport [8] are typical unbiased methods. Irradi- ance Caching [9] and Photon Mapping [3]are representative algorithms by using caching and interpolation techniques. Adaptive pixel super-sampling [7]is an in- teresting way to lower noise level. Also, filtering [11] [10] is a cheap tool to do noise removal. Re-using paths technique [1] is a powerful tool to suppress noise for MCRT renderings in an unbiased way. However, this method tends to make spike noise O. Gervasi and M. Gavrilova (Eds.): ICCSA 2007, LNCS 4706, Part II, pp. 741–750, 2007. c  Springer-Verlag Berlin Heidelberg 2007