Based on Mask
ice of Wuhan
000. Removing
Pictures.
rveying and
lging Processing
. Geomatics and
3109); pp: 733-
ing Approach in
c Engineering,
ceproduction of
iversity, Wuhan,
randbook, Third
gy of
D. Dissertation,
ing the Uneven
al of Image and
hnique Based on
^al University of
003. Use discrete
lor image
on Multispectral
(IPPR '03),
I., 2003. The
ven Elimination.
gy, 233), pp.
REMOVING SHADOWS FROM HIGH-RESOLUTION URBAN AERIAL IMAGES
BASED ON COLOR CONSTANCY
Q YE^*'H. XIE" Q. XU*
a. Department of Surveying and Geo-Informatics, Tongji University, 200092 Shanghai, China — (yeqin@tongji.edu.cn)
b.Shanghai Municipal Institute of Surveying and Mapping, 200063 Shanghai, China — (huihongxie@hotmail.com);
c.Research Center for Remote Sensing and Spatial Information, Tongji University, 200092 Shanghai, China — (xuaured@sina.com)
Commission III, ICWG III/VII
KEY WORDS: Aerial images, Color constancy, Shadow detection, Shadow removal, Shades of Gray.
ABSTRACT:
A method is explored to remove tall building shadows in true color and color infrared urban aerial images based on the theory of
color constancy. This paper first uses the specthem ratio and Otsu threshold segmentation methods to detect building shadows on
urban aerial true color and color infrared aerial images. Then, based on the shadow detection result, one of the color constancy
algorithms SoG (Shades of Gray) is used to remove the shadows in aerial images with different p values of the Minkowski norm.
Finally, the shadow removal results with different p values have been compared by brightness, contrast and average gradients. The
experiments show that the result of this method based on color constancy has a good visual effect, and different from general scene
image shadow removal, the aerial images get the best shadow removal result when p is 2. It means the two types of aerial images
should not be simply regarded as gray world images.
1. INTRODUCTION
Shadows, caused when an object obscures the light source, are
inevitable in remote sensing images. The grey value and
contrast of shadow area is obviously small in the images. Those
shadows are typically of a different color than the rest of the
image. Shadows seriously affect the visual quality, especially
for a metropolitan aerial image with a stripe shadow area caused
by high-rise buildings. As well as visual quality problems,
shadows cause difficulty in feature extraction, pattern
recognition and image matching of shadow area images,
especially for the high-resolution urban aerial images. Therefore,
shadow removal is an issue to be studied in aerial remote
sensing image processing.
There is some research on shadow detection and removal, but
most considers either the images in motion or normal nature
scene images. The commonly used methods for shadow removal
include: An integral model, removing shadow impact in log
domain of original image, thereby obtaining the image in which
the shadow effects are removed (Finlayson et al., 2002a)
(Fredembach et al, 2005); (2 Using color scale factors,
calculating shadow region scale factor, and using this factor to
get the grey value of shadow region without shadow impact
(Arbel et al., 2007). The above two methods involve complex
calculations, and the last method often performs poorly if there
are complicated textures in the image.
Shor et al propose a method for shadow detection and removal,
it needs seed in detection, and a pyramid-based restoration
process is then applied to produce a shadow-free image, image
inpainting along a thin border is finally applied to ensure a
seamless transition between the original and the recovered
regions. The result is better for normal nature scene images, but
the effect on aerial images is not known. Since it needs seed, it
is not an automatic tool (Shor et al., 2008).
Compared with a normal nature scene image, remote sensing
images have a large imaging area; every pixel's grey and color
* Corresponding author.
value are the complex composite function of solar radiation, sky
illumination, and ground reflection. Therefore the illumination
condition is much more complicated than a normal scene. Some
image enhancement algorithms, such as homomorphic filtering,
can improve the shadow impact in aerial images; but the
shadow region is a different color than the non-shadow region
when this algorithm is used in shadow removal. Some alternate
approaches have been proposed on shadow removal in remote
sensing images. Suzuki used a dynamic shadow compensation
method based on color and spatial analysis to remove shadows
in aerial images (Susuki et al., 2000); the disadvantage of this
method is it requires a probability model in advance. Finlayson
use the Retinex algorithm to remove shadows (Finlayson et al.,
2002b), but this algorithm is based on the hypothesis that the
world is grey, so it would not garner good results except for
grey areas in aerial images. The above research focuses on true
color images. Research on color infrared aerial images is
uncommon.
In this paper, our goal is to detect shadows and remove them
from true color or color infrared high-resolution urban aerial
images. The Shades of Gray (SoG) algorithm based on the color
constancy theory is chosen to remove shadows (especially
building shadows). Then some image quantitative indexes are
calculated in order to analyze the result of the shadow removal.
The optimum parameters in SoG algorithm for aerial image
shadow removal are obtained from the quantitative analysis.
2. THEORY OF COLOR CONSTANCY
Human beings have the tendency for a color to look the same
under widely different viewing conditions. Color constancy is
maintained in different lighting conditions (within a certain
range). The illumination condition is non-standard in shadow
regions. Non-standard illuminations are by definition those that
are more or less different from daylight illumination (Joint
ISO/CIE Standard ISO 11664-2:2007(E)/CIE S 014-2/E:2006).
