Material Matting

晃晃

Material Matting、

Daniel Lepage Jason Lawrence

University of Virginia

Abstract

Despite the widespread use of measured real-world materials, intu-

itive tools for editing measured reflectance datasets are still lacking.

We present a solution inspired by natural image matting and texture

synthesis to the material matting problem, which allows separat-

ing a measured spatially-varying material into simpler foreground

and background component materials and a corresponding opacity

map. We approach this problem in the context of Bayesian statistics

and introduce a new prior on materials that favors those with highly

self-similar stochastic s***cture. We describe a prototype system

that iteratively performs these separations based on small sets of

user scribbles and demonstrate multiple separations and edits.

Keywords: Appearance Models, Spatially-Varying BRDFs, Ma-

terial Separation, Texture Synthesis, Matting

1 Introduction

Modern computer graphics systems rely on accurate models of the

appearance of real-world materials. Due to the difficulty of con-

s***cting convincing appearance models by hand, they are often

measured from physical samples. However, artists still desire the

ability to edit and manipulate appearance models derived from mea-

sured data. For example, the material at the top of Figure 1, mea-

sured from the cover of an old book, has three components: scuff

marks, red ink, and an underlying golden paper. An artist may wish

to make the paper shinier, change the color of the ink, or alter the

spatial pattern of the scuffs. This paper describes a new separation

method that can assist in these tasks.

The appearance of a uniform opaque material is fully characterized

by a four-dimensional function, fr(!i; !o), called the Bidirectional

Reflectance Distribution Function, or BRDF. The BRDF is equal to

the fraction of light reflected in the direction !o due to light striking

the surface from the direction !i [Nicodemus et al. 1977]. Materi-

als whose reflectance properties vary over their surface can be de-

scribed by assigning a BRDF to every point on the surface, resulting

in a six-dimensional function Sr(~x; !i; !o), called the Spatially-

Varying BRDF or SVBRDF.

While many methods exist for editing BRDFs [Dorsey et al. 2007],

or SVBRDFs in which every point on the surface is expressed as a

linear combination of a small number of basis BRDFs [Lawrence

et al. 2006; Goldman et al. 2005], many real-world SVBRDFs are

more naturally thought of as mixtures of spatially-varying compo-

nents. The book in Figure 1, for example, is composed of several

different materials that each exhibit some degree of spatial varia-

tion, such as the different shades of red within the red stripes, or the

varying color of the golden paper.







全文请下载附件：

菜刀吻电线

非常感谢，管理员设置了需要对新回复进行审核，您的帖子通过审核后将被显示出来，现在将转入主题

奇

心中有爱，爱咋咋地

菜刀吻电线

真是不错啊

C.R.CAN

再看一看，再顶楼主

奇

佩服，好多阿 ，哈哈

奇

不错不错，收藏了

C.R.CAN

呵呵,真得不错哦!!

C.R.CAN

先顶上去，偶要高亮加精鸟！

菜刀吻电线

谢谢楼主，真是太实用了