- 最后登录
- 2018-6-29
- 注册时间
- 2011-7-1
- 阅读权限
- 20
- 积分
- 359

- 纳金币
- 335582
- 精华
- 0
|
AppGen: Interactive Material Modeling from a Single Image
Yue Dong Xin Tong Fabio Pellacini Baining Guo
Tsinghua University Microsoft Research Asia Dartmouth College Sapienza University of Rome
![]()
Abstract
We present AppGen, an interactive system for modeling materi-
als from a single image. Given a texture image of a nearly pla-
nar surface lit with directional lighting, our system models the de-
tailed spatially-varying reflectance properties (diffuse, specular and
roughness) and surface normal variations with minimal user inter-
action. We ask users to indicate global shading and reflectance in-
formation by roughly marking the image with a few user strokes,
while our system assigns reflectance properties and normals to each
pixel. We first interactively decompose the input image into the
product of a diffuse albedo map and a shading map. A two-scale
normal reconstruction algorithm is then introduced to recover the
normal variations from the shading map and preserve the geometric
features at different scales. We finally assign the specular param-
eters to each pixel guided by user strokes and the diffuse albedo.
Our system generates convincing results within minutes of interac-
tion and works well for a variety of material types that exhibit dif-
ferent reflectance and normal variations, including natural surfaces
and man-made ones.
CR Categories: I.3.7 [Computer Graphics]: Three-Dimensional
Graphics and Realism—Color, shading, shadowing, and texture;
Keywords: appearance modeling, user interaction
1 Introduction
Modeling Realistic Materials. The use of realistic materials is
necessary when rendering high-quality images. Measured materi-
als provide the highest quality datasets, but are cumbersome to use
in practice due to complex acquisition setups, lengthy measurement
times and the size of the generated data [Weyrich et al. 2009]. To-
day, the vast majority of applications use materials painstakingly
modeled by artists. Typically, artists start from a single texture
image, and use that to generate spatially-varying diffuse, specular
and roughness coefficients of an analytic reflectance model together
with a bump map to enrich the surface details. For many materials,
this process takes hours to perform, involving the use of image ma-
nipulation programs (e.g. photoshop), inverse shading tools (e.g.
CrazyBump), and shading networks in 3D software (e.g. maya).
Not only is this process cumbersome, but it often does not lead to
the highest quality materials since no robust method can be used to
easily derive detailed reflectance and normal maps from the image.
Fig. 3 shows two example materials generated by an experienced
artist in roughly one hour each, using the standard toolset that in-
cludes Photoshop and CrazyBump.
AppGen. In this paper, we present AppGen, an interactive system
for modeling material from a single image. We focus on modeling
spatially-varying reflectance (i.e. diffuse, specular and roughness
parameters) and normal variations from a texture image that is cap-
tured from a nearly planar surface lit by directional lighting. Such
images are easily found in texture collections since they are widely
used by artists when manually modeling materials. Our goal is not
to determine the exact reflectance and normals from such single
images, which is a well-known ill-posed problem. Instead, we are
interested in significantly speeding up the workflow of artists when
modeling such materials. Our key idea is that we can keep user
interaction minimal by asking the user to specify shading or re-
flectance information on a few pixels with sparse strokes, while our
algorithm efficiently infers the reflectance and normal details for
all pixels in the image. Fig. 1 shows one example of a material
modeled using our system with just a few user strokes. Note the
highly detailed, realistic look of the output material. Our experi-
enced artist was able to regenerate the example materials in Fig. 3
in a few minutes using AppGen.
全文请下载附件: |
|