![](bg1.png)
SHADOW AND SPECULAR REMOVAL BY PHOTOMETRIC
LINEARIZATION BASED ON PCA WITH OUTLIER EXCLUSION
Takahiro Mori
1
, Shinsaku Hiura
2
and Kosuke Sato
1
1
Graduate School of Engineering Sciences, Osaka University, 1-3 Machikaneyama, Toyonaka, 560-8531 Osaka, Japan
2
Graduate School of Information Sciences, Hiroshima City University, 3-4-1 Ozukahigashi, Asaminamiku, 731-3194
Hiroshima, Japan
Keywords:
Photometric Linearization, Reflection Components, Shadow Removal.
Abstract:
The photometric linearization method converts real images, including various photometric components such
as diffuse reflection, specular reflection, attached and cast shadow, into images with diffuse reflection compo-
nents only, which satisfies the Lambertian law. The conventional method(Mukaigawa et al., 2007) based on
a random sampling framework successfully achieves the task; however, it contains two problems. The first is
that the three basis images selected from the input images by the user seriously affect the linearization result
quality. The other is that it takes a long time to process the enormous number of random samples needed to
find the correct answer probabilistically. We therefore propose a novel algorithm using the PCA (principal
component analysis) method with outlier exclusion. We used knowledge of photometric phenomena for the
outlier detection and the experiments show that the method provides fast and precise linearization results.
1 INTRODUCTION
Most photometric analysis methods assume that the
input images follow the Lambertian law. It is there-
fore important to generate images with only diffuse
reflections from the input images with other photo-
metric components, such as specular reflections and
shadows.
Several methods have already been proposed
for separation of photometric components. The
dichromatic reflection model(Shafer, 1985) is often
used(Klinker et al., 1988; Sato and Ikeuchi, 1994;
Sato et al., 1997) for the separation. If the colors of
the objects are quite different from the color of the
light source, the model is very effective. However,
if the two colors are similar, the separation becomes
unstable. This method is of course not applicable for
monochromatic images.
The polarization is also useful for the separation
process. Wolff and Boult(Wolff and Boult, 1991) pro-
posed a method to separate specular reflections by
analyzing the reflected polarization, while Nayar et
al.(Nayar et al., 1993) used combined color and po-
larization clues to separate the specular reflections.
These methods, however, have a common restriction
in that they cannot handle shadows. The geometry of
the scene is useful for the analysis of specular reflec-
tions and shadows. Ikeuchi and Sato(Ikeuchi and
Sato, 1991) proposed a method to classify photomet-
ric components based on the range and brightness of
the images. A shadowed area can be distinguished us-
ing the shape of the object, but it is not easy to mea-
sure the shape of the scene even in the occluded areas.
However, there are some methods that use the
characteristics of diffuse reflection, which lies in a lin-
ear subspace. Shashua(Shashua, 1992) showed that
an image illuminated from any lighting direction can
be expressed by a linear combination of three ba-
sis images taken from different lighting directions,
and assuming a Lambertian surface and a parallel
ray. This means that an image can be perfectly ex-
pressed in a 3D subspace. Belhumeur and Krieg-
man(Belhumeur and Kriegman, 1996) showed that an
image can be expressed using the illumination cone
model, even if the image includes attached shadows.
In the illumination cone, the images are expressed by
using a linear combination of extreme rays. Georghi-
ades et al.(Georghiades et al., 2001) extended the illu-
mination cone model so that cast shadows can also be
expressed using shape reconstruction. Although any
photometric components can ideally be expressed us-
ing the illumination cone, large numbers of images
corresponding to the extreme rays are necessary.
Based on Shashua’s framework, Mukaigawa et
221
Mori T., Hiura S. and Sato K..
SHADOW AND SPECULAR REMOVAL BY PHOTOMETRIC LINEARIZATION BASED ON PCA WITH OUTLIER EXCLUSION.
DOI: 10.5220/0003817202210229
In Proceedings of the International Conference on Computer Vision Theory and Applications (VISAPP-2012), pages 221-229
ISBN: 978-989-8565-03-7
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)