DODGING IN PHOTOGRAMMETRY AND REMOTE SENSING
Zh.J. Li', Z.X. Zhang, J.Q. Zhang
College of Remote Sensing and Information Engineering, Wuhan University, 129 Luoyu Road, Wuhan, China, 430079,
zhijiang_lee(@yahoo.com.cn, zxzhang@supresoft.com.cn, jqzhang(@supresoft.com.cn
KEY WORDS: Remote Sensing, Close Range, Photography, Color, Image, Modelling, Rendering
ABSTRACT:
The images in photogrammetry and remote sensing applications are always degraded images. The degradation lies in two aspects:
geometry degradation and radiometry degradation. Aimed at the images captured in various scales, the causes of uneven illumination
and uneven color are analyzed. Then, the definition of dodging is extended as a process removes uneven illumination and uneven
color, eliminates obvious color cast, and improves the realistic representation and geometry quality effectively. However, the main
approaches for dodging currently are all merely perform the task of balancing illumination without take into account other factors.
After compared and analyzed such approaches, introduced multi-scale retinex for color rendition, and made an assumption of grey
world, a framework for color image dodging only based on image information in photogrammetry and remote sensing is proposed
and tested. The experiment indicates that such framework can process uneven lightness, uneven color, obvious color cast effectively
and restore geometry degradation in a certain degree.
1. INTRODUCTION
In the applications of photogrammetry and remote sensing, the
photos arc obtained in various carriers at various scales. The
images can be classified into remote sensing image, aero-borne
photo and close range image based on imaging scale. The
carrier of remote sensing image usually is satellite or space
shuttle. The carrier of aero-borne photo is usually the aeroplane
or other crafts. The close range image 1s captured by camera in
hand or vehicle-borne sensor, camera or video camera. In the
process of imaging, each relevant element will affect the final
image. According to the path of light propagation and
perception, such affections can be arranged as following:
(1) Mluminant. In photogrammetry and remote sensing, the
illuminant of imaging is always the sun (close range image may
be captured indoor). As to the sun, the difference of imaging
time indicates the change of light direction and the color
temperature of the illuminant. The direct affection. of the
change of light direction is that the intensity of the illumination
in the scene and the incident angle are changed. The direct
affection of changing the color temperature of the illuminant 1s
that the spectra distribution of the illuminant is changed. Thus,
the reflectance spectra of the same surface in the same scene
will be changed, which will make the image different.
(2) Propagation medium. The primary medium in
photogrammetry and remote sensing is the atmosphere. The
affections of the atmosphere on image are different in different
scales. Generally, the weather variation between multi imaging
times, the variation of the clouds and the atmosphere turbulence
etc. are primary factors, which will change the illumination
spectra on the surface in the scene.
(3) Surface characteristic in the scene. Different physical
characteristic will obtain different selective absorption of the
illumination spectra. While different geometry characteristic
will cause the interaction of multi surfaces.
(4) Imaging device. The optical system diffraction, the non-
linearity of sensors or films, the aberration of the optical system,
the relative movement of the imaging device and object in the
* zhijiang_lee@vahoo.com.cn, Tel: 86-27-87654319
scene all will cause the change of illuminant spectra and the
lightness distribution in the image.
Such affections will cause the degradation of the imaging result,
which lies in two aspects, geometry degradation and radiometry
degradation. The presentation of the former is mostly edge
blurring, noise and geometry distortion. While the presentation
of the latter is mostly uneven lightness, uneven color, and
obvious color cast.
The most intuitive approach is to build a degradation model.
However, the building of the model is various. Wu and Danaher
[Wu, 2001; Danaher, 2001] built a physical model based on the
ground reflectance. Guo [Guo, 2000] built a model based on the
radiometry characteristic of the image and imaging geometry
relation. More literatures [Tian, 1999; Wang, 2000; Chehdi,
1993; James, 1983; Zhou, 1999: Hu, 2004; Zhang, 2003a]
proposed some approaches in the view of image restoration and
image enhancement based on the image feature. However, they
can't solve all the problems simultaneously. The definition of
the dodging in such literatures is limited in dodging lightness
and color, without the realistic of color rendition. But presently,
with the development of the device and technology in image
capturing, display, and output, the request of realistic image
reproduction in the task such as image interpretation. and
analysis is enhanced at the same time. Color as an attribution is
more and more important in photogrammetry and remote
sensing, computer vision, machine vision and other areas.
Therefore, develop an approach to solve the problems above is
very important and meaningful.
Aimed at such problem, in this proposal, section II describes
the two primary directions on such topic in detail and indicates
the new definition of dodging in photogrammetry and remote
sensing. Section III relates the approach based on physical
model briefly and points out its availability. Section IV
introduces primary approaches about dodging based on the
characteristic of the image itself currently, analyzes their
function, and points out their availability. Section V proposes a
new framework about such topic. Section VI is concerned with
its application to real data and the analysis on the result.
Internc
It shoi
capture
or hyp
For re
relative
of radi
betwee
actual.
the uni
and re
Based
correct
atmosp
the dis
affecti
elemen
restore
A lot c
include
atmosp
radiom
and so
correct
image |
and is
althoug
accurat
quantit
applica
and fe
correct:
Additic
image «
visual i
etc.. B«
sensing
atmosp
the ima
take int
is diffic
of the :
scales t
applica
and is a
Therefc
process
illumin:
to buik
informa
With tl
device
realistic
dodging
extende
dodging
obvious
degrada
psychol
convent
