stanbul 2004
Xf the scene.
aries from ()
the colour or
ir dimension
:d and green
s for RGB to
re (Carper et
is study the
stribution of
it each. DN.
the spatial
images may
ver, spatial
r example, a
'O dominant
howengerdt,
. The RGB-
r composites
erefore their
stograms of
iat would be
n the RGB-
monitor the
mage. Hue-
| of image
ator of the
Koutsias et
r for whole
histograms:
DNs.
ly the DNs
percentual
/ average).
Comission,
gram; there
of median.
ommon.
used out of
)9%-part of
Ns between
.995, 0,99,
ing amount
5 DNs.
percentual
ole image.
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B1. Istanbul 2004
3. MATERIALS AND METHODS
3.1 Materials
The investigation was made in close co-operation with three
Finnish image producers, Ministry of Agriculture and Forestry
(MAF), two Forestry Centres and the Forestry Development
Centre Tapio. In the following, the image producers are named
Company A, Company B and Company C.
The study of histograms was based on 41 scanned CIR images
from Company A and 2818 histograms collected with
HISTOQC-program from image servers of Company A and
Company B. The scanning process was done using either 10 or
12 bits per channel and images were transformed to 8 bits per
channel afterwards. Summary of the histogram material used is
presented in table 1. Scanners used are presented in table 2.
can be used either interactively or from the command line.
HISTOQC calculates the statistics described on chapter 2.2.
The black borders of images, resulted from surrounding film
clear base, were excluded from the calculations of histograms.
Statistic-files were imported to Excel and several charts and
diagrams were created and analysed.
Experimental image enhancement was made with Adobe
PhotoShop 7.0 —software. Effects of these enhancements to
image histograms were followed using HISTOQC.
4. RESULTS
4.1 Histograms
Different types of films produced clearly distinguishable image
histograms. In all images investigated, RGB-histograms were
typically unimodal, i.e. they had a single peak. In colour films
Table 1. Histograms used.
Scanner [Model Bits [Owner
SI Zeiss SCAI 10 | Company A
S3 Leica Geosystems DSW600 12 | Company A
S3a |Leica Geosystems DSW600 10 | Company B
S3b |Leica Geosystems DSW600 12. | Company B
Table 2. Scanners.
For the study of the tone tuning procedures and their influence
on radiometric quality, 20 tuned CIR ortho images were
received from three Finnish image producers and two Forestry
Centres. Company A, Company B and Company C delivered
three types of images: the original orthophoto, one tuned to be
used in forest interpretation and one tuned to be used in land
parcel interpretation. Forestry Centres delivered images tuned
to be used in forest interpretation. Company A also delivered
41 scanned CIR images (Kodak 1443 CIR-negative film, Leica
Geosystems DSW600 12bit scanner). 40 of the images were
taken in different parts of Finland (10 from Central Finland, 10
from Northern Karelia, 10 from Northern Ostrobothnia, 10
from Lapland) and in different seasons (from each area: 5 in
the beginning of summer 2002 and 5 in the middle or end of
the summer 2002). In addition, one image from southern
Finland taken on the summer 2003 was delivered.
3.2 Methods
Preliminary studies of histograms were made with Matlab 6.5
and Erdas Imagine 8.5 —softwares. Based on these results, a
program named HISTOQC was coded with C++ -language for
collecting histograms and statistics of scanned aeria! images. It
runs in Windows-based operating systems in DOS-prompt and
Source Amount [Country [Film Scanner {Resolution
Company A 5 SWE [Agfa X-100 (colour-neg.) > 14pm
Company A 48 FIN {Avi Chr 200 (colour-pos.) SI 14 pm Copurpositivefim Colour-negative-film
Company A 18 FIN [Avi Chr 200 (colour-pos.) S2 14 um i
CompanyA | 277 FIN [Kodak 1443 (CIR-neg.) SI 14 jm iu se
Company À 557 FIN | [Kodak 1443 (CIR-neg.) S2 14 pum ET genre
Company À 6 LTU |Kodak 1443 (CIR-neg.) SI l4 um op i |^ 20000 k [^
Company A 131 LTU [Kodak 1443 (CIR-neg.) S2 14 pm 3000000 1 3000000 3 [3
Company A 2 SWE [Kodak 2444 (colour-neg.) SI 14 um 2000000 H 2000000 | E
Company A 92 SWE [Kodak 2444 (colour-neg.) S2. 14 um ol | 1000000 { /
Company B 96 FIN Kodak 1443 (CIR-pos.) S3a 20um ol | o
Company B 911 FIN Kodak 1443 (CIR-pos.) S3b 20 pm | ut run |[- s i
Company B 608 FIN Kodak 1443 (CIR-pos.) S3b 14um CIR-positive -film CIR-negative-film
6000000 6000000
5000000 4
5000000
4000000 j| | 4000000 1——
3000000 —8 3000000
2000000 4-:——— 2000000
1000000 1000000
04
Figure 2. Histograms of different types of films.
this peak was often situated on the darker half and in CIR-
films on the brighter half of histograms. The latter could be
explained by the strong near-infrared and green reflectance of
vegetation. In CIR-films bimodal histograms may appear if
there are large water areas on the image. Example histograms
of different types of films are presented in figure 2.
4.0 QC of Scanned images
Data from the Company A image server were calculated with
the first version of HISTOQC-program, so only R, G and B —
histograms and their statistics were gathered. Data from the
Company B image server included also the L-histograms.
In the analysis the histogram material was classified to image-
detailed (Figure 3), flight-detailed (Figure 5) and film- and
scanner-type-detailed data (Figure 4). According to studies of
these histograms, the most important statistics for the QC were
efficiency, saturation (0% & 255%) and 99%-efficiency. In the
ideal situation, efficiency of all histograms in 8-bit data would
be 256 and there would be only a small amount of 0 and 255
DNs. Because the scanning process is such that several images
are scanned using the same parameters, some compromises
have to be accepted, and full utilization of histogram cannot be
obtained. Also the averages and standard deviations of
