Texture Mapping Based on Coefficient Map and
Basic Texture Map
Huijian Han
School of Computer Science & Technology, ShanDong Economic University, Jinan ,China
Shandong Prov. Key Lab of Digital Media Technology, Jinan ,China
Email: hhj@sdie.edu.cn
Hengwu Li
School of Computer Science & Technology, ShanDong Economic University, Jinan ,China
Email: smxlhw@sina.com
Abstract—Texture mapping may give the impression of
geometric details in a model using an image. But texture
always was captured under special light condition. If the
lighting in virtual environment is different from the texture
image, the result of rendering will be incorrect and
unrealistic. This paper proposes an image-based method
that requires basic texture map and coefficient maps to
interpolate light effective. This method uses a quadratic
multinomial to fit the reflection model. The coefficients of
quadratic multinomial will be gained from BTFs and are
stored as coefficient maps. A picture is taken under well-
proportioned environment light as a basic texture map,
which the chromaticity is saved. The method can
reconstruct the surface color under varying lighting
conditions and represent the variation in surface color for
each texel independently. Coefficient map and basic texture
map make texture mapping become more realistic, simple
and convenient.
Index Terms—texture mapping, coefficient map, basic
texture map, iamge-based
I. INTRODUCTION
In the reality scene, there are three basic structures in
the geometry, namely macrostructure, mesostructure and
microstructure. The macrostructure with the certain
geometry shape can be seen by the eyes, such as building,
furniture shape etc. Mesostructure with quite small
geometry shape still can be seen, for instance orange’s
skin. Microstructure is the micro unit of surface can’t be
seen. The Microstructure affects optical quality such as
light scattering. The mesostructure causes visual effect
[1] like roughness, self-shadows, occlusions, inter-
reflection and subsurface scatting etc., which is an
important factor that we get the realistic object surface
with rich detail [2]. Mesostructures are typically rendered
using techniques such as bump mapping [3], horizon
mapping [4] or displacement mapping [5]. Mesostructure
and microstructure decide the optical quality and the
detail visual quality of object surface.
Using traditional texture, for example a realistic image,
may realistically increase the model’s geometry detail by
mapping texture to the surface of object. However,
because the texture is generally get by taking photographs
under some special viewpoint position and specific
lighting condition. When this texture is mapped to the 3-
D object surface, the lighting condition in the virtual
scene is not considered. If the lighting in the virtual
environment is consistent with the lighting which the
texture was captured under, the reality is the most strong.
Contrarily, the result of rendering will appear incorrect
and unrealistic. Bump mapping provides basic shading,
which perturbs mesh normals to match those of the fine
geometric detail, but not shadowing, occlusion, and
silhouettes. Introducing variations in the surface normals
causes the lighting method to render the surface as though
it had local surface variations instead of just a smooth
surface. Bump maps can be either hand modeled or, more
typically, calculated procedurally. But it is still difficult
to create a bump map base on real pictures.
This paper offers an image-based method for
representing various lighting effect, which is suitable for
the diffuse and specular reflection object. The pictures are
captured under fixed viewpoint and under kinds of
illuminations condition. We choose a polynomial model
to describe the variation of each texel’s luminance. The
polynomial coefficients can be stored as maps for basic
texture. The basic texture and its coefficients maps can be
mapped to the object simultaneously. This method can
reconstruct the texture’s luminance and color under
varying lighting conditions.
II.
RELATED WORK
With very simple geometries, texture and bump
mapping yield good results for simple materials, but for
more complex materials we need the ability to change the
appearance for varying light and viewing conditions.
Early approaches simulated a single BRDF for the whole
material [6]. Kautz and McCool [7] approximated the
BRDF by two functions, whose results are stored in
textures and were combined by the graphics hardware.
These methods, which improved by [8] [9] and [10], lit
Corresponding author: Huijian Han
Email: hhj@sdie.edu.cn; hanhuijian@yahoo.com.cn
Adress : No.7366 ERHUAN Road, Jinan city, China, 250014
412 JOURNAL OF MULTIMEDIA, VOL. 4, NO. 6, DECEMBER 2009
doi:10.4304/jmm.4.6.412-418