Image Analysis
Learning Objectives: -
By the end of this topic, you will be
able to:
Ø
Realize the different
elements of visual interpretations required.
Ø
Tell how digital image
processing is done.
Ø
Tell what is image
preprocessing and enhancement.
Ø
Learn about the image
classification.
Ø Differentiate between supervised
classification and unsupervised classification.
Introduction:
à
Image analysis is the
extraction of meaningful information from images; mainly from digital images by
means of digital image processing techniques.
à
Many image processing
and analysis techniques have been developed to aid the interpretation of remote
sensing images and to extract as much information as possible from the images.
à The choice of specific techniques or algorithms
to use depends on the goals of each individual project.
Elements of Visual Interpretation:-
There
are 8 elements of Visual Interpretation. They are:
1.Tone 5.Texture
2.Shape 6.Shadow
3.Size 7.Size
Factor/ Topological Location
4.Pattern 8.Association
Size:
à
Size of objects in an
image is a function of scale.
à
It is important to
assess the size of a target relative to other objects in a scene, as well as
the absolute size, to aid in the interpretation of that target.
à
A quick approximation of
target size can direct interpretation to an appropriate result more quickly.
à For example, if an interpreter had to distinguish
zones of land use, and had identified an area with a number of buildings in it,
large buildings such as factories or warehouses would suggest commercial property,
whereas small buildings would indicate residential use.
Shape:
1. Shape refers to the general form, structure, or outline of individual
objects.
2.
Shape can be a very distinctive clue for interpretation.
3.
Straight edge shapes typically represent urban or agricultural (field) targets, while natural
features, such as forest edges,
are generally more irregular in shape, except where man has created a road or
clear cuts.
4. Farm or crop land irrigated by rotating sprinkler systems would appear as
circular shapes.
Shadow:-
1. Shadow is also helpful in interpretation as it may provide an idea of the
profile and relative height of a target or targets which may make identification
easier
2.However, shadows can also reduce or eliminate interpretation in their area of
influence, since targets within shadows are much less (or not at all)
discernible from their surroundings
3.Shadow
is also useful for enhancing or identifying topography and landforms, particularly
in radar imagery
Tone:-
1. Tone refers to the relative brightness or colour of objects in an image.
2. Generally, tone is the fundamental element for distinguishing between
different targets or features.
3.Variations in tone also allow the elements of shape, texture, and pattern of
objects to be distinguished.
Texture:
1.Texture refers to the arrangement and frequency of tonal variation in
particular areas of an image.
2.Rough textures would consist of a mottled tone where the grey levels change
abruptly in a small area, whereas smooth textures would have very little tonal
variation.
3.Smooth textures are most often the result of uniform, even surfaces, such as
fields, asphalt, or grasslands.
4.A target with a rough surface and irregular structure, such as a forest
canopy,
results in a rough textured appearance.
5.Texture is one of the most important elements for distinguishing features in
radar imagery.
Pattern:
-
1.
Pattern refers to the
spatial arrangement of visibly discernible objects.
2. Typically, an orderly repetition of similar
tones and textures will produce a distinctive and ultimately recognizable pattern.
3. Orchards with evenly
spaced trees, and urban streets with regularly spaced houses are good examples
of pattern.
Site Factor/Topological Location:-
1. Relative elevation or specific location of objects can be helpful to
identify certain features
2. For example, sudden appearance or disappearance of vegetation is a good clue
to
the underlying soil type or drainage conditions
Association:
1. Association takes into account the relationship between other recognizable
objects or features
in proximity to the target of interest.
2. The identification of features that one would expect to associate with other
features may
provide information to facilitate identification.
3. For example, commercial properties may be associated with proximity to major
transportation routes, whereas residential areas would be associated with
schools, playgrounds, and sports fields.
Digital Image Processing:-
Digital Image Processing (DIP) is a technique which involves manipulation of
digital image to extract information. When satellite images are being
manipulated in such a manner, this technique is also referred to as satellite image
processing.
It
involves combination of software-based image processing tools. The whole
process of Digital Image Processing can be classified into three parts:
1) Digital Image Pre-Processing
2) Digital Image Enhancement
3) Digital Image classification
Geometric Correction Methods:-
1. The information extracted from remotely sensed images is integrated with map
data in a geographical information system.
2.The transformation of a remotely sensed image into a map with a scale and projection
properties is called geometric correction.
3. To correct sensor data, both internal and external errors must be
determined.
4. Geometric rectification of the imagery resamples or changes the pixel grid
to fit that of a map projection or another reference image.
5. This becomes especially important when scene to scene comparisons of individual
pixels in applications such as change detection are being sought.
1. The primary function of remote sensing data quality evaluation is to monitor
the performance of the sensors
2.The performance of the sensors is continuously monitored by applying
radiometric correction models on digital image data sets
3.Radiometric Correction Methods include:
1. Computation of Radiance (L)
2. Computation of Reflectance
3. Cosmetic Operation
4. Random Noise Removal
1.Computation of Radiance:-
1.Radiance is a measure of the radiant energy given out by an object and picked
up by a remote sensor.
2.Spectral radiance (L) is defined as the energy within a wavelength band
radiated by a unit area pre solid angle of measurement.
Radiance (L.) = (Dn/D max) (Lmax -Lmin) +Lmax
Dn = digital value of a pixel from
the computer-compatible tape (CCT)
Dmax = maximum digital number recorded on the CCT
Lmux = maximum radiance measured at detector
saturation in mW
Lmin = minimum radiance measured at detector saturation in mW
,
2.Computation of Radiance:-
Reflectance is an energy ratio which is a function of radiance
Reflectance =
(Radiance)/ E sina
3.Cosmetic Operations:-
1.The corrections involved in the cosmetic operations are the correction of
digital images containing either partially or entirely missing scan lines and
the correction of images because of destripping of imagery,
2. Due to the variation in the sensitivity of the detectors, the irradiance of
the object may differ.
3. The line dropout or missing scan line is usually overcome by replacing the
zero value by the mean values of the pixels of the previous and the following
line.
4.Random Noise Removal:-
1. Image noise is any unwanted disturbance in image data that is due to
limitations in the sensing and data recording process.
2. The random noise problems referred to as spiky are characterized by non-systematic
variations in gray levels from pixel to pixel.
Atmospheric Correction Methods:-
1.According to Rayleigh scattering, the effect of scattering is inversely
proportional to the fourth power of the energy.
2. The bias is the amount of offset for each spectral band.
3.Bias can be determined by regressing the visible band vs infrared bands
4.To correct the scattering, firstly identify some areas and then the
brightness values of all these features from each band is extracted.
5.The gray value of the visible band is plotted against corresponding values at
the same pixel location in the infrared band.
6. The plot will result in a scatter diagram, following the regression analysis
regression line is to be fitted.
7. If the data is free from atmospheric scattering, the best fitting line
should pass
through the origin.
Image Enhancement:-
1.Low sensitivity of the detectors, weak signal of the objects present on the
earth surface, similar reflectance of different objects and environmental
conditions at the time of recording are the major causes of low contrast of the
image.
2.Image enhancement is done to amplify these slight differences for betterment
of the image scene.
3.Image enhancement is defined as mathematical operations that are to be
applied to digital remote input data to improve the visual appearance of an
image for better interpretability or subsequent digital analysis.
4.Contrast stretching
5.Spatial filtering
6.Edge enhancement
7.Linear data transformations
8.Stretch the contrast to enhance
9.Features interested
10.Linear stretch (with or without saturation)
11.Sinusoidal or sine
12.Non-Linear
For example, equalization, standard deviation, etc.
Spatial Filtering:-
1.The neighborhood
2.The image profile
3.Numerical filters
4.low-pass filters
5.high-pass filters
1.The neighborhood: -
2.The image profile: -
3.Numerical filters: -
4.low-pass filters: -
5.high-pass filters:-
High and Low Frequency
Changes: -
6.Edge enhancement:
Linear Data Transformation: -
1. The individual bands are often observed to be
highly correlated or redundant.
2. Two mathematical transformation techniques
are often used to minimize this spectral redundancy
3. Principal component analysis (PCA)
4. u canonical analysis (CA)
Principal
Components Analysis:-
1. Compute a set of new, transformed variables
(components), with each component largely independent of the others
(uncorrelated).
2. The components represent a set of mutually
orthogonal and independent axes in a n-dimensional space.
3. The first new axis contains the highest
percentage of the total variance or scatter.
PC Images: -
1. The digital enhancement techniques available
in remote sensing are contrast stretching enhancement, rationing, linear
combinations, principal component analysis and spatial filtering.
2. The image enhancement techniques are broadly
classified as point operations and local operations.
3. Point operations modify the values of each
pixel in the image dataset independently.
4.Local operations modify the values of each
pixel in the context of the pixel value surrounding it.
5. In Linear contrast stretch, the digital
number (DN) value in the lower end of the original histogram is assigned to digital
number zero i.e., extremely black and value at the higher end is assigned to
extremely white.
6. The intermediate values are interpolated
between 0 and 225 by following a linear
relationship Y = a + bx
7.Exponential contrast enhancement is also considered
as a non-linear contrast enhancement.
8. The grey values in the input image transform
to the gray values in the output image
9. Historical equalization is widely used for
contrast manipulation in digital image processing because it is very simple for
implementation.
10. Here, the original histogram has been
readjusted to produce a uniform population density of pixels along the
horizontal grey value axis.
Image Classification:
-
1. Image classification is the process of
creating a meaningful digital.
2. Thematic map from a image data set
information extraction).
֍
Supervised
classification: Classes from known cover types.
֍ Unsupervised classification: Classes by
algorithms that search the data for similar pixels.
The
process of image classification: -
Supervised Classification: -
1.Classification methods that relay on use of
training patterns are called supervised classification methods.
2.A supervised classification algorithm requires
a training sample for each class.
3.The training samples are representative of the
known classes of interest to the analyst.
Training
stage: The analyst identifies
representative training areas and develops numerical descriptions of the
spectral signatures of each land cover type of interest in the scene.
The classification stage: Each pixel in the image
data set is categorized into the land cover class it mostly resembles. If the
pixel is insufficiently similar to any training data set it is usually labelled
'unknown'.
The output stage: Three typical forms of
output products are thematic maps, tables and digital data files. The output of
image classification becomes input data for GIS for spatial analysis of the terrain.
1.Training
class selection (training areas/classes)
2.Generating statistical parameters (spectral
signatures) of training classes
3.Data classification
4.Evaluation and refinement
Supervised Spectral Classification:
Common
Classifiers:
1. Parallelepiped
2. Minimum distance to mean
3. Maximum likelihood
4. Parallelepiped Approach
5. Simple method
6. Makes few assumptions about character of the
classes
7. Minimum distance to mean
8. Find mean value of pixels of training sets in
n-dimensional space
9. All pixels in image classified according to
the class mean to which they are closest
10. Maximum likelihood
11. All regions of n-dimensional space are
classified
Clustering:
-
Clustering
Process: -
Classification
Procedures: -