International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004
Weekly
External quality control rr report Contractor
: A a "Audi —#iPhotographing re > 2
— Checking of quality repdrts ilm development |
— Acceptance control
Quality report
«—« tet
Quality
control
“Result |
Audit : ;
ie Scanning
«-—Qualitvirepot
+ ICP
rej |
Acceptd nce, I— der
control T.
Scanned
images
Figure 1. QC-strategy of image production.
The CIR orthophotos are also extensively used in the forestry
applications. Orthophotos used in forest planning are
radiometrically enhanced to assist forest interpretation. Most of
Forestry Centres receive images from companies as tuned all
the way, while some do most of the enhancement process
themselves. The radiometry of these images have been
estimated only visually, criteria have been subjective and
varied depending on the person. Because of this, the QC of
radiometrically-enhanced orthophotos has been a difficult task
for both to the image producer and customer.
The main objectives of this study were to develop automatic
procedures for 100%-histogram control of scanned images and
to develop an objective way for judging radiometry of enhanced
orthophotos. The existing FLPIS orthophoto QC-system will be
improved based on the results.
24. RADIOMETRY OF IMAGE
2.1 Digital image composition and colour spaces
À normal digital colour image consists of three channels: red
(R), green (G) and blue (B) and one pixel is composition of
these three (R, G, B). At the moment 24-bit colour images are
used, they contain one 8-bit value (0-255) for each channel.
Hence, there "are up to 256^3 = 16777216 different
combinations of red, green and blue Digital numbers (DN) that
one colour pixel may have. In the future with new digital
cameras and image processing software also 10-tol4-bit data
per channel may be used, film-scanners can already produce up
to 12-bit per channel images.
In addition to R, G and B-channels, several composition
channels based on these three can be calculated from image.
Luminosity describes the panchromatic information of the
image and in Adobe Photoshop it is calculated with a formula
(1) (Kosaka, 2000).
L 2 0.30*R 4 0.59*G -- 0.11*B (1)
Intensity (I), hue (H) and saturation (S) colour space is
advantageous to RGB in that it presents colours more nearly
perceived by the human eye. Intensity is comparable to
250
luminosity and describes the overall brightness of the scene.
Saturation represents the purity of colour and varies from 0
(grey) to 1 (pure colour). Hue is representative of the colour or
dominant wavelength of the pixel. It is a circular dimension
. and varies from 0 at the blue midpoint through red and green
back to the blue midpoint at 360. Several formulas for RGB to
IHS —transformation can be found from the literature (Carper er
al., 1990; Gonzalez and Woods, 1992), but in this study the
formula given by Conrac (1985) was used.
2.2 Histogram
The image histogram describes the statistical distribution of
image pixels in terms of the number of pixels at each DN.
Particularly, it contains no information about the spatial
distribution of those pixels; two visually different images may
have identical histograms. Sometimes, however, spatial
information can be inferred from the histogram. For example, a
strongly bimodal histogram usually indicates two dominant
materials in the scene, such as land and water (Schowengerdt,
1997).
The most widely used histograms are R, G, and B. The RGB-
model is applied for producing three-channel colour composites
on colour monitors, scanners and other devices, therefore their
histograms are the most efficient to calculate. Histograms of
other composition channels may give information that would be
hard or even impossible to interpret straight from the RGB-
histograms. Luminosity-histogram can be used to monitor the
number of totally black or white pixels on the image. Hue-
histogram can be used to monitor the effect of image
enhancement operations, and it is a better indicator of the
overall tone of the image than RGB-histograms (Koutsias ef
al., 2000). A histogram can be calculated either for whole
image or for a smaller part of it.
Several statistical quantities can be calculated from histograms:
— Average. Describes the mean value of the DNs.
— Standard deviation. Describes, how widely the DNs
are spread around the average.
— Coefficient of Variation. Describes the percentual
deviation of the DNs (=standard deviation / average).
— EC Coefficient of Variation (European Comission,
2004). (=standard deviation / 256).
— Median. Describes the center of the histogram; there
are an equal amount of DNs on both sides of median.
— Mode. Describes the DN that is the most common.
— Efficiency. Describes how many of DNs is used out of
256. Opposite to unused DNs.
- 99%-efficiency. Describes the width of 99%-part of
the histogram (= tail 0.995 - tail 0.005).
— Minimum. The smallest DN used.
- Maximum. The largest DN used.
— Unused Center. The number of unused DNs between
minimum and maximum.
=: i. Tails, 0.001,.0.008,.0.01,.0.05.: 0.999, 0.995, .0,99,
0.95. Describes the place were corresponding amount
of histogram is on the left side.
= 0 & 255. Describes the amount of 0 and 255 DNs.
- 0% & 255% (saturation). Describes the percentual
amount of ( and 255 DNs compared to whole image.
31
Th
Fin
Ce)
Co
Th
fro
HE
Co
ch:
SITZ] =
e
Tolololololoio!
rem
