Spectral analysis for automatic 3-D texture generation

Abstract

Texture is an important notion for achieving a high degree of realism in image synthesis. 2-D texture mapping methods are well known techniques used in computer graphics with some classical problems. The two major drawbacks of texture mapping are texture discontinuity and texture distortion in some object regions. A more recent and efficient approach to resolve the 2-D texture mapping problems is solid texturing ( 3-D texturing). Although interesting, solid texturing is complex owing to the problems that the existing 3-D texture generation methods raise. In fact, there is no reliable way to define directly the desired type of texture with these methods. So, the existing 3-D texture generation methods remain rather experimental. In addition, aliasing is another important problem to solve with solid texturing because the conventional and common antialiasing approaches of texture mapping become very expensive in this case. In this paper we present a new, powerful and simple 3-D texture generation method. It is based on a spectral analysis of a 2-D texture (digitized image) of the same type. The 2-D spectral analysis, using a Fourier transform, is used to extract some significant parameters of the texture necessary for a 3-D texture generation. This method allows us to define easily the desired type of texture for an automatic 3-D texture generation. It is simpler and more efficient than the existing 3-D texture generation methods. It can produce a realistic simulation of a wide range of natural textures such as wood, marble, cloud, water waves, etc. In addition, a simple and direct texture antialiasing can be naturally provided by this method.