Elsevier

Computers & Graphics

Volume 28, Issue 5, October 2004, Pages 747-756
Computers & Graphics

Technical Section
Illumination-dependent texture

https://doi.org/10.1016/j.cag.2004.06.012Get rights and content

Abstract

In this paper, we propose an image-based texture model called illumination-dependent texture (IDT), for realistically rendering objects covered with three-dimensional(3D) texture under low-frequency illumination. A framework of IDT is presented, which includes acquisition, analysis, synthesis and rendering. After sampling 3D texture images, we analyze samples based on spherical harmonics. Then the IDT samples could be synthesized on arbitrary surface. Before rendering, we take preprocesses for IDT coordinate transformation and illumination analysis. Then a simple rendering process could be performed in real time under low-frequency natural environment illumination, or simple lighting (parallel lighting or point lights). Finally, we demonstrate several IDT rendering results with real-world and synthetic texture samples.

Introduction

Texture mapping is an effective technique for rendering realistic image. Since the BTF model was introduced by Dana [1], many research works such as [2], [3], [4], have been focused on the microstructure modeling for texture surface. As a 6D function, BTF model takes into account the influence of both view direction and lighting direction on three-dimensional(3D) texture appearance. In order to exhibit high-quality visual effects, previous BTF models usually take a large number of sampling images and get the texel color from these samples directly, while much storage and computation efforts are consumed. Polynomial texture maps (PTM) is a simplified model proposed by Malzbender [5]. This texture model finely simulates surface detail as lighting moves relative to the object. Although the influence of view direction is neglected, this light-dependent representation can render most 3D detail of planar object surface under parallel lighting. However it does not introduce synthesis method for PTM in [5].

Inspired by PTM model we propose a light-dependent 3D texture model called illumination-dependent texture (IDT). Our goal is realistic rendering of textured objects under distant natural illumination. Our work shows that the light-dependent representation could also approximate the textured surface in low-frequency lighting situation very well. Fig. 1 shows an IDT demo, and Fig. 2 compares a surface texture and a surface IDT. The IDT model extends the bi-quadratic polynomial model for PTM in [5] to a complete spherical harmonics (SH) representation. And in fact the PTM model is an incomplete 2-order SH model. The texture surfaces with various microgeometry details could be modeled by SH expansions in appropriate order, instead of a fixed expression in [5]. Most importantly, this representation can easily cooperate with SH-based environment maps and pre-computed radiance transfer, while the results in [5], [6], [7] can only be rendered under simple lighting. Another advantage of IDT model is that the real-world texture samples acquisition is very convenient compared with [7], which requires large amount of sampling data and complex camera calibration and image registration work.

We also introduce the IDT synthesis technique for arbitrary surface to transplant the 3D texture samples onto virtual objects. Finally we render the textured objects under environment illumination in real-time.

The main contributions of this paper are (1) we propose a SH-based representation for 3D texture model (IDT), (2) develop an IDT synthesis method for arbitrary surface, and (3) introduce an efficient rendering method for IDT-covered objects under low-frequency illumination.

The rest of this paper is organized as follows. After a brief overview of related work in Section 2, we introduce the IDT representation in Section 3. Section 4 discusses the IDT synthesis technique, and the rendering process is described in Section 5. Results and conclusion are reported in Section 6.

Section snippets

Environment maps and BRDF analysis

The SH-based environment maps analysis proposed by Ramamoorthi [8] renders diffuse objects under distant illumination. The lighting and surface normals are expressed as 2-order SH functions. The irradiance can be calculated very efficiently by the dot product of these two coefficient vectors. Ramamoorthi [9] extends the above method to render objects with complex BRDFs, and proposes the spherical harmonic reflection maps. Similar work was done by Kautz [10]. Latta [11] and Mccool [12] apply the

Overview

IDT is an image-based technique that requires no modeling or rendering of complex geometry for surface microstructure. This technique could not only represent the surface normals, but also exhibits the self-shadowing, sub-surface scattering, and inter-reflections. The overview of IDT framework is depicted in Fig. 3. The input data is a set of sampling images of texture surface under different illumination, while the camera is fixed. Then the texel reflectivity function is represented by the SH

IDT synthesis on surface

Texture synthesis on surface is a challenging task. For IDT synthesis on surface, we take a simple but effective approach. The synthesis pipeline includes: mesh simplification; vector field generation; texture patch searching and blending; and mesh subdivision (optional). This method can also be applied for normal texture synthesis.

Rendering

The IDT rendering can be performed efficiently under simple lighting condition and low-frequency environment illumination. We first take some preprocesses for coordinate system transformation and illumination analysis. Then the surface texture in the atlases is generated on the fly. Finally the model is rendered with these texture atlases by graphics hardware.

Results and conclusion

We take both real-world and synthetic texture samples in our experiments. For each IDT sample, about 50 images are captured at 250*250 resolution in sampling process. For most demos, the 2 order SH expansions are used, and the 3D models are simplified to about 2k triangles. The synthesis process for each model takes approximately half an hour without acceleration. At most two 1024*1024 texture atlases are used. All preprocesses for rendering spend a few minutes. We render the result images at

Acknowledgements

Thanks to Paul Debevec for his light probes (http://www.debevec.org). This paper is supported partially by 973 project (No. 2002CB312100) and NFSC Grants of No. 60033010 and No. 60021201.

References (31)

  • K. Dana et al.

    Reflectance and texture of real-world surfaces

    ACM Transactions on Graphics

    (1999)
  • Dana K, Nayar S. Correlation model for 3D texture. Proceedings of the International Conference on Computer Vision,...
  • Cula O, Dana K. Compact representation of bidirectional texture functions. In: Proceedings of IEEE Conference on...
  • T. Leung et al.

    Representing and recognizing the visual appearance of materials using three-dimensional textons

    International Journal of Computer Vision

    (2001)
  • Malzbender T, Gelb D, Wolters H. Polynomial texture maps. Proceedings of SIGGRAPH’01, 2001. p....
  • Liu X, Yu Y, Shum H. Synthesizing bidirectional texture functions for real-world surfaces. Proceedings of SIGGRAPH’01,...
  • Tong X, Zhang J, Liu L, Wang X, Guo B, Shum H. Synthesis of bidirectional texture functions on arbitrary surfaces....
  • Ramamoorthi R, Hanrahan P. An efficient representation for irradiance environment maps. Proceedings of SIGGRAPH’01,...
  • R. Ramamoorthi et al.

    On the relationship between radiance and irradiancedetermining the illumination from images of a convex Lambertian object

    Journal of Optical Society of America A

    (2001)
  • Kautz J, Snyder J, Sloan P. Fast arbitrary BRDF shading for low-frequency lighting using spherical harmonics. EGWR...
  • Latta L, Kolb A. Homomorphic factorization of BRDF-based lighting computation. Proceedings of SIGGRAPH’02, 2002. p....
  • Mccool M, Ang J, Ahmad A. Homomorphic factorization of BRDFs for high-performance rendering. Proceedings of...
  • Sloan P, Kautz J, Snyder J. Precomputed radiance transfer for real-time rendering in dynamic low-frequency lighting...
  • Ramamoorthi R, Hanrahan P. Frequency space environment map rendering. Proceedings of SIGGRAPH’02, 2002. p....
  • Blinn J. Computer display of curved surfaces. PhD Thesis, University of Utah,...
  • Cited by (0)

    View full text