Image enhancement techniques are used to improve the appearance of digital imagery for visual interpretation. There are linear and non-linear enhancement methods. Linear contrast stretching uniformly expands a narrow range of pixel values in an image histogram to utilize the full tonal range. Histogram equalization assigns a wider range of values to frequently occurring portions of the histogram. Spatial filtering can highlight or suppress specific spatial frequencies to enhance or reduce image texture. Principal components analysis transforms image bands into uncorrelated components that capture decreasing amounts of variance, reducing dimensionality while retaining most information. Vegetation indices like NDVI use ratios of infrared and red bands to identify vegetated areas.
Image Enhancement Techniques for Visual Interpretation
1. Image Enhancement
Applied to more effectively display or
record the image data for subsequent
visual interpretation.
linear enhancements
non-linear enhancements
2.
3. The objective of the second group of image processing
functions grouped under the term of image enhancement,
is solely to improve the appearance of the imagery to
assist in visual interpretation and analysis. Examples of
enhancement functions include contrast stretching to
increase the tonal distinction between various features in
a scene, and spatial filtering to enhance (or suppress)
specific spatial patterns in an image.
4. The key to understanding contrast enhancements is
to understand the concept of an image histogram.
A histogram is a graphical representation of
the brightness values that comprise an image.
The brightness values (i.e. 0-255) are displayed
along the x-axis of the graph. The frequency of
occurrence of each of these values in the image
is shown on the y-axis.
5. Linear contrast stretch.
This involves identifying lower and upper bounds from the
histogram (usually the minimum and maximum brightness
values in the image) and applying a transfn to stretch this range
to fill the full range.
In our example, the
min. value (occupied
by actual data) the
histogram is 84 and
The max. value is 153.
These 70 levels occupy
less than of the full 256
levels available.
A linear Stretch
uniformly expands
This small range to
Cover the full range of values from 0 to 255.
6. This enhances the contrast in the image with light toned
areas appearing lighter and dark areas appearing darker,
making visual interpretation much easier. This graphic
illustrates the increase in contrast in an image before (left) and
after (right) Linear contrast stretch.
7. A uniform distribution of the input range of values across the full range
may not always be an appropriate enhancement, particularly if the input
range is not uniformly distributed.
In this case, a histogram-equalized stretch may be better.
This stretch assigns more display values (range) to the frequently
occurring portions of the histogram.
In this way, the detail in these
areas will be better enhanced
relative to those areas of the
original histogram where
values occur less frequently.
In other cases, it may be
desirable to enhance the
contrast in only a specific
portion of the histogram.
8. For example, suppose we have an image of the mouth of a river,
and the water portions of the image occupy the digital values
from 40 to 76 out of the entire image histogram. If we wished to
enhance the detail in the water, perhaps to see variations in
sediment load, we could stretch only that small portion of the
histogram represented by the water (40 to 76) to the full
grey level range (0 to 255).
All pixels below or
above these values
would be assigned to 0
and 255, respectively,
and the detail in these
areas would be lost.
However, the detail in
the water would be
greatly
enhanced.
9. Spatial filtering
Encompasses another set of digital processing functions which
are used to enhance the appearance of an image. Spatial
filters are designed to highlight or suppress specific features in
an image based on their spatial frequency. Spatial frequency
is related to the concept of image texture,It refers to the
frequency of the variations in tone that appear in an image.
10. A low-pass filter is designed to emphasize larger,
homogeneous areas of similar tone and reduce the smaller
detail in an image. Thus, low-pass filters generally serve to
smooth the appearance of an image. Average and median
filters, often used for radar imagery, (low-pass filters.)
High-pass filters do the opposite and serve to sharpen
the appearance of fine detail in an image. One implementation
of a high-pass filter first applies a low-pass filter to an image
and then subtracts the result from the original, leaving behind
only the high spatial frequency information.
11.
12. Multi-band operations
To enhance or extract features from satellite images which cannot be
clearly detected in a single band, you can use the spectral information of
the object recorded in multiple bands.
These images may be separate spectral bands from a single multi spectral data
set, or they may be individual bands from data sets that have been recorded at
different dates or using different sensors. The operations of addition,
subtraction, multiplication and division, are performed on two or more coregistered images of the same geographical area. This section deals with multiband operations. The following operations will be treated:
- The
use of ratio images to reduce topographic effects.
- Vegetation indexes, some of which are more complex than
ratio's only.
- Multi-band statistics.
- Principal components analysis.
- Image algebra, and;
- Image fusion.
13. Image transformations
When a satellite passes over an area with relief, it records both shaded and
sunlit areas. These variations in scene illumination conditions are illustrated in
the figure. A red silt stone bed shows outcrops on both the sunlit and the
shadowed side of a ridge. The observed DNs are substantially lower on the
shaded side compared to the sunlit areas. This makes it difficult to follow the silt
stone bed around the ridge.
In the individual Landsat-TM bands 3 and 4,
the DNs of the silt stone are lower in the
shaded than in the sunlit areas. However,
the ratio values are nearly identical,
irrespective of illumination conditions.
Hence, a ratioed image of the scene
effectively compensates for the brightness
variation, caused by the differences in the
topography and emphasizes by the color
content of the data.
14. Image Interpretation
Vegetation Indices
Different bands of a multispectral image may be combined to
accentuate the vegetated areas. One such combination is the
ratio of the near-infrared band to the red band. This ratio is
known as the Ratio Vegetation Index (RVI)
RVI = NIR/Red
Since vegetation has high NIR reflectance but low red
reflectance, vegetated areas will have higher RVI values
compared to non-vegetated areas. Another commonly used
vegetation index is the Normalised Difference Vegetation
Index (NDVI) computed by
NDVI = (NIR - Red)/(NIR + Red)
15. Various mathematical combinations of satellite bands, have
been found to be sensitive indicators of the presence and
condition of green vegetation. These band combinations are
thus referred to as vegetation indices. Two such indices are the
simple vegetation index (VI) and the normalized difference
vegetation index (NDVI).
NDVI = (NIR – Red) / (NIR + Red)
Both are based on the reflectance properties of vegetated areas
as compared to clouds,water and snow on the one hand, and
rocks and bare soil on the other.
Vegetated areas have a relatively high reflection in the near-infrared and
a low reflection in the visible range of the spectrum. Clouds, water and
snow have larger visual than near-infrared reflectance. Rock and bare soil
have similar reflectance in both spectral regions.
The effect of calculating VI or the NDVI is clearly demonstrated in next table.
16. Table: Reflectance versus ratio values
TM Band 3 TM Band 4 VI
NDVI
Green
Vegetation
21
142
121
0.74
Water
21
12
-9
-0.27
Bare soil
123
125
2
0.01
17.
18. Vegetation maps are produced by generating a normalized
difference vegetation index from a infrared image and then
doing a vegetation classification. Color infrared photographs
collect information in the green, red and near infrared light
reflectance spectrum. Green vegetation reflects very strongly
in the near infrared light range and therefore infrared images
can detect stress in many crops before it is visible with the
naked eye
19. The Normalized Difference Vegetation Index (NDVI) is
used to separate green vegetation from the background
soil brightness. It is the difference between the near
infrared and red reflectance normalized over the sum of
these bands.
20. NDVI = (IR-Red)/(IR+Red)
These NDVI maps can then be classified into vegetation categories and
displayed as a vegetation maps with different colors representing different
levels of vegetation. In the map on the left browns and yellow represent bare
soil and shades of green represent vegetation, darker greens are stronger
vegetation.
21. 5.3.3 Principal
components analysis
Another method (which is not spatial and applied in many fields), called principal
components analysis (PCA), can be applied to compact the redundant data into
fewer layers. Principal component analysis can be used to transform a set of image
bands, as that the new layers (also called components) are not correlated with one
another.
These new components are a linear combination of the original
bands. Because of this, each component carries new information. The
components are ordered in terms of the amount of variance explained, the
first two or three components will carry most of the real information of the
original data set, while the later components describe only the minor
variations (sometimes only noise).
Therefore, only by keeping the first few components most of the information
is kept. These components can be used to generate an RGB color composite, in
which component 1 is displayed in red, component 2 and 3 in green and blue
respectively.
Such an image contains more information than any combination of the three original
spectral bands.
22. To perform PCA, the axis of the spectral space are rotated,
the new axis are parallel to the axis of the ellipse (see
figure). The length and the direction of the widest transect of
the ellipse are calculated. The transect which corresponds to
the major (longest) axis of the ellipse, is called the first
principal component of the data. The direction of the first
principal component is the first eigenvector, and the variance
is given by the first eigenvalue.
A new axis of the spectral space is defined by the first
principal component. The points in the scatter plot are now
given new coordinates, which correspond to this new axis.
Since in spectral space, the coordinates of the points are the
pixel values, new pixel values are derived and stored in the
newly created first principal component.
23. The second principal component is the widest transect of the ellipse that is
orthogonal (perpendicular) to the first principal component. As such, PC2 describes
the largest amount of variance that has not yet been described by PC1. In a two
dimensional space, PC2 corresponds to the minor axis of the ellipse. In ndimensions there are n principal components and each new component is
consisting of the widest transect which is orthogonal to the previous components.
24. The output of a PCA contains the following tables:
· Variance/covariance matrix between the original bands
(variables).
· Correlation matrix between the original bands.
· Component’s eigenvalues – the amount of variance
explained by each of the new bands (%variance =
eigenvalue /
S[eigenvalue]).
· Eigenvectors – the parameters for the linear combination
of the new bands for an inverse transformation, back
to the original bands.
· Component’s factor loadings (factor pattern matrix) –
factors with a high loading parameter with an original
band, have a high correlation with it.
25. Linear Combination
For the Morro Bay TM scene there are 7 spectral bands. Thus each
pixel has 7 values. The pixel in row i, column j of the image is a
vector:
x(i,j,1) x(i,j,2) x(i,j,3) x(i,j,4) x(i,j,5) x(i,j,6) x(i,j,7)
x(i,j,1) is the value of band 1 in row i, column j, x(i,j,2) is the value of
band 2 in row i, column j, etc.
A linear combination of these values, to calculate the first Principal
Component, would look like:
This multiplication and addition is carried out for each of the picture
elements, pixels, in the image. The Principal Components Analysis is
the calculation of the values of the set of vectors a and then the
multiplication of the image data by them to get the projections of the
data points onto the Principal Components.
26.
27. Image Processing and Analysis
3. Classification
• Bands of a single image are used to identify and separate spectral
signatures of landscape features.
• Ordination and other statistical techniques are used to “cluster” pixels of
similar spectral signatures in a theoretical space.
• The maximum likelihood classifier is most often used.
• Each cluster is then assigned to a category and applied to the image to
create a classified image.
• The resulting classified image can now be used and interpreted as a
map.
•The resulting classified image will have errors! Accuracy assessment is
critical. Maps created by image classification should report an estimate of
accuracy.
28. Image Processing and Analysis
3. Classification
Black
Box
Transformation / Clustering
Maximum Likelihood Classifier
Band 4
Spectral Signatures
Band 3
Band 2
Band 1
Classified Image (Map)
29. Modeling / GIS
Remote Sensing Images as GIS layers
For est Type
Geology
Land Ownership
Sub strate and
Owner ship Mask
Elevation
Regression
Model
Soil Dept h
Soil pH
Predicted Density of Reeses Buttercup
ote sensing data (raw or processed) is most powerful when incorporated into a GIS.