Landsat Color Composite Notes

SFR406 – Spring 2015
Formation and Interpretation of Color Composite Images
Introduction
One sees color images collected by earth orbiting satellites in popular magazines, in movies
and television shows; however, few viewers have any understanding of what they are
looking at. Many of these images were recorded by Landsat or other multispectral scanners.
These images are usually “color composite” images (as described in the following sections).
Color composite images are often interpreted visually, sometimes to compliment data
analysis in remote sensing studies. For example, one can learn how to interpret general
vegetation types from a summer color composite image to determine where to go to visit or
sample cover type attributes in the field.
Color composite images combined with a digital elevation model (DEM) can depict some
interesting perspectives of the landscape (Figure). Visual interpretation of time-series
Landsat color composites has been used as a method to support accuracy assessment of
forest or land cover change images (Cohen et al., 1998; Sader et al., 2003). In regions where
vegetation or land cover maps do not exist or are out of date, visual interpretation of
Landsat color composites has been employed as the method to develop the vegetation maps
for large mapping areas (FAO, 1993).
Landsat color composite image draped on a Digital Elevation Model
Digital and color composite images can be interpreted visually as in aerial photo
interpretation or digitally as in digital image processing performed with computer
programs. Color composite images can be interpreted visually using many of the same PI
principles used in aerial photo interpretation. Students well- grounded in PI principles and
related concepts will be able to adapt their interpretation skills to digital images,
particularly the interpretation of satellite color composite images formed on a computer
monitor or printed/plotted on paper.
Fundamentals Concepts
The art of interpreting a 3 band color composite image is aided by the following:
1) Field knowledge and distribution of cover types throughout the study area (e.g.,
forest types and plant ecology, topography, and terrain analysis.
2) Familiarity with the multispectral reflectance characteristics and DN levels for
different forest and land cover type.
3) Familiarity with additive color theory.
4) Knowledge of PI principles and PI experience.
Mechanics of a Color Composite Image
If three different wavebands are displayed simultaneously on the computer monitor, a color
composite image will result. The color “write functions” or color guns in the monitor are
represented by primary colors - red, green and blue (RGB). Two colors in combination form
complementary colors (R + G = yellow; R + B = magenta; G + B = cyan, all 3 primaries
combined are white and lack of all three form no color (black). Each image represents 8 bits
and three 8 bit images in the composite represents over 16 million possible colors.
The additive primaries: red, green, and blue
•
Characteristics:
–
Equal proportions of the three additive primaries combine to
form white light.
The strength of the composite colors formed are controlled by the relative brightness or DN
in each of the 3 wavebands coupled with R, G, and B. For example, a blue pixel represents
high reflectance and thus a high DN coupled with the B color plane, and the wavebands
coupled with R and G have very low, or no DN values. For example deep, clear water may
appear blue on a true color composite because there might be higher reflectance in the
shortest waveband (visible blue) and most of the energy at the other wavelengths are
mostly absorbed, thus little to none reflected (and hence low DN for wavelength couple with
R and G color gun in the computer monitor). Clouds that are not moisture laden are white
on all color composites, because they reflect highly in all visible and reflected infrared
bands on any composite image combination.
The figure above depicts a generic color table interpretation chart to understand the
relation of waveband DN and additive color theory (primary and complementary colors that
result). Once the student understands the principles of wavelength dependent reflectance of
earth surface features, image interpretation principles, and additive color theory, one can
develop the skill to interpret a color composite image collected by satellites or other
sensors.
With these skills and understanding, one should be able to view a satellite color composite
image collected virtually anywhere in the world and know what the general cover types are,
without ever being there. One must keep in mind that all remote sensing work needs to be
supported by field observations, but where it is not possible, these interpretation skills can
be invaluable for a general overview of the area.
References
Cohen, W.B., M. Fiorella, J. Gray, E. Helmer, and K. Anderson. 1998. An efficient and accurate
method for mapping forest clearcuts in the Pacific Northwest using Landsat imagery.
Photogrammetric Engineering and Remote Sensing 64(4):293-300.
[FAO] Food and Agricultural Organization. 1993. Forest Resources Assessment, 1990,
Tropical Countries. FAO Forestry Paper 112. Rome: Food and Agricultural Organization
of the United Nations.
Sader, S.A., Bertrand, M., & Wilson, E.H. 2003. Satellite change detection of forest harvest
patterns on an industrial forest landscape, Forest Science, 49(3), 341-353.
True Color, CIR and False Color Composite Images using Landsat 8 OLI
Interpreting a True Color Composite
To create a true color composite, the three visible bands available on Landsat are coupled
with the primary colors in the computer monitor (R = visible red, G = visible green, and B =
visible blue). Other names for this composite are “normal” or “natural” color.
Landsat -8 OLI
RGB 432 “true”
color composite
Vegetation types
are variations of
green; urban and
inert surfaces are
white; water is dark
blue; bog/wetlands
are brownish.
There is a cloud in
the upper left
corner of the image
with thin clouds
and haze trending
in the southeast
direction from the
main cloud.
This composite image will have similar color to true or normal color aerial photos and the
way humans see color. Healthy vegetation is green, water is dark blue to black (depending
on depth, turbidity, algae content, etc). Landsat is capable of displaying a true color
composite but most other medium spatial resolution, multispectral scanners on orbiting
commercial satellites cannot, because they do not contain all three visible wavebands. Many
interpreters prefer true color, because colors look natural to our eyes. Perhaps one
disadvantage of true color is that some different earth surface features have similar low
reflectance in the visible wavelengths making it difficult to interpret subtle difference
among vegetation types, for example.
Interpreting a Color Infrared Composite
To create a color infrared composite image, two visible bands and the near infrared band
are combined (R = NIR, G = visible red, B = visible green). This color pattern simulates the
same color patterns as seen on color infrared photos although a different color process
(subtractive primary colors) is involved with aerial photos and film. This is also called a
false color composite (this is from a human perspective because it is not true color the way
we see it). All composites that do not form true color can be categorized as false color. On
the color infrared composite, healthy vegetation is reddish to magenta. Softwood are darker
red-brown and hardwood are brighter red-magenta. Water is black.
Landsat 8 OLI
RGB- 543 “Color
Infrared” composite
image
This composite
simulates the color
of a color infrared
aerial photo and
can be interpreted
using the same
logic.
Vegetation types
are variations of
magenta; urban
features and bare
field are cyan.
This composite
contains the near
infrared waveband
and therefore
vegetation types are
better distinguished
compared to a true
color composite
False Color Composite OLI RGB- 564 and OLI RGB- 654
This false color composite image is one of the best for showing color differences between
vegetation types (e.g., hardwood, softwood, wetland types).
The TM wavebands and primary colors in the computer monitor are coupled as follows: R =
NIR, G = Mid IR, B = visible red. Hardwoods are orange to yellow (high R (NIR), mediumhigh G (Mid IR), and low B because the visible red is a chlorophyll absorption band. In the
TM 654 composite, the near infrared and mid-infrared are rearranged with the red and
green color guns so that vegetation appears green.
Landsat OLI
RGB- 564 false color
composite.
This composite
contains the visible
red and both the
near infrared and
mid infrared bands.
Although the colors
are not natural to
human eyes, the 564
band combination is
arguable the best for
distinguishing
different forest and
vegetation types in
the northeastern U.S.
S= Softwood
H= Hardwood
Landsat OLI
RGB- 654 color
composite
The false color
composite contains
the same 3 bands as
the previous example
however the (near
infrared) band (5) is
coupled with the
green color gun.
Because vegetation
reflects higher in the
near infrared than
the other two bands,
the green color
dominates. Some
interpreters prefer
this combination
because it makes the
vegetation look more
natural.