ering, 
ntional 
digital 
wn that 
devices 
canners 
device 
N 
icies of 
ry and 
ed the 
of lens 
points 
acy of 
imiting 
g them 
ulation 
ig from 
applied 
ed this 
a test 
| on 3rd 
E color 
IT8.7/2 
momial 
values 
rs, and 
imetric 
device 
or this 
ire and 
a new 
1apping 
without 
etween 
istimuli 
values. This method is particularly useful either when it is 
crucial to transform a particular color with great accuracy or 
when the calibration data is incomplete. 
Ostromoukhov et al (1994) performed 3D-transformations 
between RGB and CIE XYZ color spaces in order to calibrate 
electronic display systems. 
Mohd and Kirby (1997) tested geometric accuracy of desktop 
publishing scanners by using a calibrated grid plate. They 
applied affine transformation and polynomial regression 
between scanned Cartesian coordinates and the calibrated ones. 
They found out that 2nd and 3rd order polynomials are suitable 
for geometric improvement of scanned images. Higher order 
polynomials are no more useable. The order of polynomial 
depends on the scanner used. 
Noriega ct al (2001) determined some properties of scanners by 
using negative and positive density measurements on the 
scanned RGB values and the CIE XYZ color space values. 
They used standard observer color matching functions to 
transform the density values into CIE XYZ values. They 
concluded that the inaccuracies of the colorimetric values by 
scanner depend on the properties of the device and the related 
color management system. 
In the models based on polynomial regression, including more 
samples in this part of the colorant space could reduce the errors 
at the gamut boundary. Common test targets such as the ISO 
12640 and 12641 have only a limited number of patches at the 
gamut boundary. Therefore, Green (2000) proposes a new 
target for defining media gamut boundaries. He used 2nd order 
polynomial regression in his application. 
Kang (1997) mentioned that many scientists have successfully 
applied regression method for transformation of scanned RGB 
values to colorimetric values. He determined polynomial 
equations of different order that convert the RGB values to CIE 
XYZ and CIE Lab color space values. He used IT8.7/2 
calibration card and Sharp JX 450 scanner. He compared the 
results obtained by different order of polynomials and different 
standard illuminants. 
Berns and Shyu (1994 and 1995) proposed a color mixing 
method based on theories of Beer-Bouger and Kubelka-Munk 
and scanner signals. They also applied polynomial regression. 
Hardeberg (1999) developed methods based on linear 
regression, polynomial regressions of 2nd and 3rd order. He 
applied these methods to AGFA Arcus 2 scanner with AGFA 
IT8.7/2 calibration card. 
Yilmaz (2002) applied conformal, affine, projective and 
polynomial transformations for reducing color inaccuracies. He 
examined five different scanners. He found out that the 
polynomial regression of 3rd order delivers the best results in 
general. 
3. SCANNERS 
À scanner is an electronic device that captures the image of any 
object and converts it into a computer format. It includes an 
array of cells that are optically sensitive. These cells called 
CCD (Charge Coupled Device) are connected with a detector 
that measures the intensity of coming light and converts it to 
electrical signals. The reflecting light from the object is 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004 
projected to the detector via an optical system. The electrical 
signals generated by the detector are then converted to digital 
information. Any of the information represents a pixel, and its 
size is a number of bits that are saved per pixel. If the number 
of bits per pixel increases, the quality of image becomes better. 
According to usage areas, scanners can be classified as follows 
(Baltsavias & Bill, 1994): 
e Photogrammetric scanners 
e Scanners of large documents 
e  Microdensitometers 
* Desktop publishing scanners 
e — Other scanners such as scanners of 3-D objects, slide 
scanners, text document scanners, multiple purpose 
scanners, etc. 
According to working principle two types are distinguished: 
Drum scanners and flat bed scanners. Drum scanners can 
process large format documents and delivers accurate images. 
The most common type of flat bed scanners is desktop 
publishing scanners. 
The reliability of scanners depends on the certain tests. Four 
types can be performed (Yakar, 2002). 
l. Geometric accuracy: With this test geometric 
distortions are investigated. Calibration targets with 
maximal positional error of 2-3 micron are required. 
2. Geometric resolution: This is the test of optical 
scanning system. A variety of calibration targets are used. 
3.  Radiometric accuracy: This is the test of the 
sensibility of scanner to gray tones. 
4. Color accuracy: In this test the original colors and the 
scanned colors are compared. Calibration cards or targets 
with standardized colors are required. 
After these tests it is decided if the tested device is suitable for 
certain tasks. 
4. COLOR CALIBRATION CARDS 
Color calibration cards used for scanner calibration and test for 
color accuracy have been known as Q-60 calibration cards. 
They are mainly used in printing industry, photography, and for 
calibration of monitors, printers and scanners. The main 
purpose is to have best color outputs on reflection and 
transmission materials. They are also used for color calibration 
of scanners. Most of the cards, which are based on the Kodak’s 
Q-60 color calibration card, are produced according to ANSI 
and ISO standards. 
Kodak’s Q-60 calibration cards, having all colors of the CIE 
Lab color space, are in accordance with the ANSI IT8 7.1 and 
[T8 7.2 and ISO 12641 standards. In general there are two 
types: IT8 7.1 (transmission) and IT8 7.2 (reflection). IT8 7.1 is 
printed on “Ektachrome Professional” films, with the size of 
4x5 inch. IT8 7.2 is printed on “Ektachrome Professional” 
cards. Its dimensions are 5x7 inch. Both consist of 240 color 
parts, 24 gray parts and a human face. The colorimetric values 
of the cards are given by the manufacturer in a text file. In these 
files, CIE XYZ color space values, CIE Lab color space values, 
Sx, Sy, Sz (standard deviations of CIE XYZ values), average 
density (D) and SD (standard deviation of average density) are 
given. In this study Kodak's IT8 7.2-1993 2001:02 calibration 
card was used.