Efficient Visualization of Parameter Space for Rendering Translucent Objects

Translucent objects have a characteristic appearance due to a phenomena called subsurface scattering of light. In computer graphics, many methods have been proposed for rendering the translucent objects realistically. However, it is often difficult to generate an image with the desired appearance; the user must repeatedly generate images with different parameter settings, which is tedious and time-consuming. Our goal in this paper is reducing such user’s burden for the parameter tuning. We propose an efficient visualization method of the parameter space for the subsurface scattering so that the user can explore the space interactively. Our basic idea is creating a database consisting of images rendered with randomly sampled parameters and generating images with arbitrary parameters by interpolation. However the parameter space is usually high-dimensional, the costs for the database storage and the interpolation calculation are both expensive. We address this problem by combining the principal component analysis (PCA) and the radial basis functions (RBF). Using our method, the images with different parameters are calculated quickly, allowing her/him to efficiently find the optimal parameters that produce the desired appearance.