International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004 Intei
k,=GrevyRIR
kei. ur (2)
c — GrevB /
k, = GreyB/B
Where, GreyR, GreyG, GreyB are the value of grey in the
scene. Ravers Gavers Baver are averages of each channel.
aver
5.2 MRSCR
The Multi-scale Retinex (MSR) [Rahman, 1996a] is a
generalization of the single-scale retinex (SSR) [Jobson, 1996],
which, in turn, is based upon the last version of Land's
center/surround retinex. À later version, the MSRCR, combines
the dynamic range compression and color constancy of the
MSR with a color ‘restoration’ filter that provides excellent
color rendition [Rahman, 1998]. The MSRCR has been tested
on a very large suite of images. The Retinex theory assumes
that human vision is based on three retinal-cortical systems,
each processing the low, middle and high frequency of the
visible spectrum independently [Marini, 2000]. The general
form of the MSRCR can be summarized by the following
equation: [Rahman, 1996b]
S
R,, (x,y) = F, (x, y) 2, w, (log[7, (x y)]-
s=l
log[ 1, (x, y) * M, (x, y)b. is] 2, 0 Ne (3)
Where R. is the ith band of the MSRCR output, S is the
number of scales being used, w, is the weight of the scale, I; is
the ith band of the input image, and N is the number of bands in
the input image. The surround function M, is defined by
M (x,y) = K explo? /(? € y^]
where * eis the standard deviation of the sth surround function,
and. [[K exp[oz /G? y^ dxdy 71.
Fi(x,y) are the color restoration function defined by
Ex op E
S nox)
n=l
However, from the Tab.l, it's clear that the MSRCR can't
balance the lightness of the image, although it can obtain a
suitable dynamic range, wonderful color rendition and keep
geometry information effectively. Then the lightness balancing
is needed.
5.3 Lightness balancing
To avoid decreasing the contrast of the image, the methods
listed in Tab.1 are abnegated. Here the lightness balancing can
be accomplished just by gain/offset rectification in little
windows of the image. It can keep the tone information in detail
effectively, although the window size is always be given firstly.
6. RESULT AND CONCLUSION
Two origin images above are processed by the framework
proposed in Fig.11. Compared with images processed by
traditional dodging methods and merely MSRCR, the
difference is clear.
However, since the MSRCR can process image in several DJ:
scales, the lightness balancing should be united into the form of Perf
MSRCR. What's more, as a method based on retinex theory, on I
which is describing the human perception on color as an
important theory of computational color constancy, the new Kan
version should include such functions. It’s also the next Tecl
direction of us in the work for image re-rendition. 364|
Kob
and
Edw
The
Lan
Ame
Li >
Grot
Rem
ACKNOCKLEDGEMENTS D. N
Colc
Thanks for the supporting from Natural Science Fund of P. R. 18
China (No. 40171081) and Surveying and Mapping Fund of
State Bureau of Surveying and Mapping of P. R. China (No. Mila
2001-02-03). proc
Thoi
REFERENCES Pres
Blinn Jim, Newell Martin, 1976. Texture and Reflection in T.Pe
Computer Generated Images. In: Communications of ACM, [mag
19(10): 542-547. 37.8
Chehdi, K., 1993. An Optimal Image Enhancement Method By Z.Rz
Decisional Criteria. In: Proc of ICSP '93, pp. 981-984. Retii
IEE]
Chuang Yung-Yu, Zongker D.E. Hindorff Jetal, 2000. et
Environment Matting Extensions: Towards Higher Accuracy Zug
and Real-time Capture. In: SJGGRAPH2000, pp.121-130. Mult
Com
T. Danaher, X. Wu and N. Campbell, 2001. Bi-directional Scie
Reflectance Distribution Function — Approaches to Digi
Radiometric Calibration of Landsat TM imagery. In: :
IGARSS 2001 conference, pp.2654-2657. Z.R
of tt
Fournier A, Gunawan A, Romanzin C, 1993. Common Proc
Illumination between Real and Computer Generated Scenes. In:
Proceedings of the Graphics Interface'93, pp.254-262. Sher
Surv
F. Gasparini, R. Schettini, 2004. Color Balancing of Digital J.Co
Photos Using Simple Image Statistics. Pattern Recognition, 37
(6), pp. 1201-1217. Shur
Imag
Guo Xiao-Fang, 2000. Radiometry Correction of Airborne SPIE
Imaging Spectrometer Data. Journal of Image and Graphics,
5(1), pp.16-20. Taka
Renc
Hu Qing-wu;-Li-Qing-quan, 2004. Image Restoration Based on Com
Mask Technique. Geomatics and Information Seience-of Wuhan
University, 29(4), pp. 319-323. E Tian
base
James, D.F., Robert, A.S., 1983. Spatially Variant Contrast Petri
Enhancement Using Range Modification. SP/E, 22(3), pp. 378- 9(4),
381.
: Tian
R. Jaroszhiewicz, 2003. Fast Extraction of BRDFs and Material Corr
Maps from Images. Master Thesis. Jour,
922
