IAPRS & SIS, Vol.34, Part 7, *Resource and Environmental Monitoring", Hyderabad, India,2002
both the needs of the photogrammetric as well as of the remote
sensing users. This encourages us to speak about a
“photogrammetric remote sensing" concept.
2.2 Coloration of High-Resolution Pan-chromatic Pixels
Of interest to us is the manner in which color is being assigned
to each high-resolution pixel. Traditionally this was not an
important topic since multi-spectral scanning provided each
pixel with multiple color values. However, with color being
scanned at one geometric resolution and by one sensor, and the
panchromatic information being collected at a higher geometric
resolution and with a separate sensor, the issue emerges of how
to "color" the black-and-white panchromatic pixels. A standard
"coloration" scheme has emerged in which the panchromatic
pixels represent the high geometric information and thus the
"intensity" band in a color space, whereas the color pixels
represent the hue and saturation or similar such bands. More
generally, coloration can be accomplished in various ways as
discussed below.
2.3 Sensing Color
2.4.1 Three CCDs
Each of three CCDs has its own color filter for red-green-blue
(RGB), and each CCD images with a different perspective (see
Figure 1).
Figure 1: Three images are being produced, one by each of the 3 CCDs,
and represent the RGB-channels. These images must be co-registered.
The problem with this approach is the need to overcome the
mis-registration between the RGB-channels. Registration
typically is being accomplished via matches of the R-G and G-
B-channels.
2.3.2 CCD with Color Beam Splitter.
Figure 2 illustrates the use of a beam splitter to produce 3
images with the same geometry. The issue is the reduced
amount of light arriving at the sensor.
2.3.3 Single CCD with Spinning Wheel Disk Filter
Figure 3 suggests the spinning wheel to obtain color in 3
separate images, taken sequentially. This limits the geometric
quality since the images are taken with separate exterior
orientations during flight.
2.3.4 Bayer Pattern CCD for Color Information
As proposed by Bayer (1976), a single area array CCD is being
used with a filter in front of every CCD-element. Figure 4
Figure 2: Beam splitter to obtain color in one simultaneous step (from
www.howstuff works.com). There exists a single geometry.
Registration matching is not needed.
Figure 3: The so-called "spinning wheel" to sequentially create red,
green and blue component images, taken at slightly different times,
since the filters must be rotated into place in front of the lens.(from
www.howstuffworks.com).
J DTV DOD
ONO BD
2078 | 201 0| 20
OO OK
DDD DD
Figure 4: The Bayer pattern is obtained by placing red, green and blue
filters over each element of a CCD as shown. Green is being favored
over red and blue.
Figure 5: A simulated illustration of a Bayer pattern image. Left is a
conventional color image with 81 x 81 pixels. To the right is the
computed Bayer pattern image with alternating r, g and b pixels in the
pattern shown in Figure 4. That is the raw output of the color CCD.
Reconstruction of the final color image is called demosaicking.
illustrates the basic concept and shows that green has twice the
number of red- and blue-pixels. Figure 5 is a simulation on 81 x
81 pixels of a color image using this basic idea. At issue is the
creation of the color image from the sensed pattern. This is
denoted as “demosaicking” (Adams, 1995, 1997).
233K
Foveon
6). W
www. fc
"The re
three l
silicon
light pe
first ful
first im
every p.
The lar
This mi
Figure 6
Foveon,
2.4 A:
We are
scheme
the gec
by 400
2000 x
pixels
2000 t
resolut
resolut
resolut
filter, c
Geome
respon:
basic f
quality
color :
colorat
along
based
panchr