ER 3707 Digital Image Processing July 2016 Exam – Model Answers

Question 1
(i). DIP comprises the following four basic steps:
(a) Image correction/restoration: Image data recorded by sensors on a satellite or aircraft contain
errors related to geometry and brightness values of the pixels. These errors are corrected using
suitable mathematical models, which are either definite or statistical models.
(b) Image enhancement: Image enhancement is the modification of image, by changing the pixel
brightness values, to improve its visual impact. Image enhancement techniques are performed by
deriving the new brightness value for a pixel either from its existing value or from the brightness
values of a set of surrounding pixels.
(c) Image transformation: The multi-spectral character of image data allows it to be spectrally
transformed to a new set of image components or bands with a purpose to get some information
more evident or to preserve the essential information content of the image (for a given application),
with a reduced number of transformed dimensions. The pixel values of the new components are
related to the original set of spectral bands via a linear operation.
(d) Image classification: The overall objective of image classification procedures is to
automatically categorize all pixels in an image into land cover [Marks 10]

(ii). DIP data formats

The most common formats are:
- Band Interleaved by Pixel (BIP),
- Band Interleaved by Line (BIL), and
- Band Sequential (BSQ).
To appreciate the data formats, consider a hypothetical remote sensing dataset containing just nine
pixels obtained over the interface between land and water.
The Image consists of three bands (band 1=green; band 2=red; band 3=near-infrared).
The brightness value (BVi,j,k) row, column, and band notation is provided (figure 1, a).
Band Interleaved by Pixel (BIP) Format
The BIP format places the brightness values in n bands associated with each pixel in the dataset in
sequential order [e.g., for a dataset containing three bands the format for the first pixel in the matrix
(1,1) is 1,1,1; 1,1,2; 1,1,3]. The brightness values for pixel (1,2) are then placed in the dataset (e.g.,
1,2,1; 1,2,2; and 1,2,3), and so on. An end-of-file (EOF) marker is placed at the end of the dataset
(figure 1,b).
Band Interleaved by Line (BIL) Format
The BIL format creates a file that places the brightness values in n bands associated with each line
in the dataset in sequential order.
For example, if are three bands in the dataset, all of the pixels in line 1, band 1, line 1, band 2, and
then line 1, band 3. An EOF marker is placed at the end of the dataset (figure 1,c).
Band Sequential (BSQ) Formats
The BSQ format places all of the individual pixel values in each band in a separate and unique file.
Each band has its own beginning header record and EOF marker (figure 1,d)

[Marks 30]
(iii). Optimum Index Factor (OIF)
Optimum Index Factor (OIF) is one of the most common statistical methods which were applied
in order to designate the most favorable three band combinations that has the information with the
least amount of duplication. It is based on the total variance within bands and correlation
coefficient between bands.

[Marks 20]

(iv). Elements of Visual Image Interpretations

Eight fundamental parameters or elements are used in the interpretation of remote sensing images
or photographs. These are tone or color, texture, size, shape, pattern, shadow, site and association.
In some cases, a single such element is alone sufficient for successful identification; in others, the
use of several elements will be required.
• Tone
Tone is the relative brightness of grey level on black and white image or color/F.C.C image. Tone
is the measure of the intensity of the reflected or emitted radiation of the objects of the terrain.
Lower reflected objects appear relatively dark and higher reflected objects appear bright.
• Shape
Shape refers to the general form, configuration or outline of an individual object. Shape is one of
the most important single factors for recognizing object from an image. Generally regular shapes,
squares, rectangles, circles are signs of man-made objects, e.g., buildings, roads, and cultivated
fields, whereas irregular shapes, with no distinct geometrical pattern are signs of a natural
environment, e.g., a river, forest.
• Size
Size of objects on images must be considered in the context of the image scale or resolution. It is
important to assess the size of a target relative to other objects in the scene, as well as the absolute
size, to aid in the interpretation of that target. A quick approximation of target size can make direct
interpretation to an appropriate result more quickly. The most measured parameters are length,
width, perimeter, area, and occasionally volume
 • Pattern
Pattern refers to the spatial arrangement of the objects. Objects both natural and manmade have a
pattern which aids in their recognition. The repetition of certain general form or relationship in
tones and texture creates a pattern, which is characteristic of this element in image interpretation.
• Texture
Texture refers to the frequency of tonal variation in an image. Texture is produced by an aggregate
unit of features which may be too small to be clearly discerned individually on the image. It
depends on shape, size, pattern and shadow of terrain features. Texture is always scale or resolution
dependent. Same reflected objects may have difference in texture helps in their identification.
• Shadow
Shadow is a helpful element in image interpretation. It also creates difficulties for some objects in
their identification in the image. Knowing the time of photography, we can estimate the solar
elevation/illumination, which helps in height estimation of objects. The outline or shape of a
shadow affords an impression of the profile view of objects.
• Association
Association refers to the occurrence of certain features in relation to others objects in the imagery.
In urban area a smooth vegetation pattern generally refers to a playground or grass land not
agricultural land
• Site:
Site refers to topographic or geographic location. It is also an important element in image
interpretation when objects are not clearly identified using the previous the elements
[Marks 20]
Question 2
(i).
Scattering is the redirection of electromagnetic energy by suspended particles in the atmosphere.
The type and amount of scattering that occurs depends on the size of the particles and the
wavelength of the energy. Atmospheric constituents such as gas molecules, smoke and dust, water
vapor may cause for the scattering.The three main constituents which absorb radiation are Ozone,
Carbon Dioxide, and Water Vapor.

Ozone serves to absorb the harmful (to most living things) ultraviolet radiation from the sun.
Without this protective layer in the atmosphere our skin would burn when exposed to sunlight.

Carbon Dioxide absorbs in the far infrared portion of the spectrum which is related to thermal
heating and results in a 'greenhouse' effect.
Water Vapor absorbs energy depending upon its location and concentration, and forms a primary
component of the Earth's climatic system.

[Marks 20]
(ii).

There are three main types of scattering that impact incoming solar radiation:

• Rayleigh Scatter
• Mie Scatter
• Non-Selective Scatter

Rayleigh scatter occurs when radiation (light) interacts with molecules and particles in the
atmosphere that are smallerin diameter that the wavelength of the incoming radiation. Shorter
wavelengths are more readily scattered that longer wavelengths. Light at shorter wavelengths (blue
and violet) are scattered by small particles that include NO2 and O2. Since blue light is at the short
wavelength end of the visible spectrum, it is more strongly scattered in the atmosphere than long
wavelength red light. This results in the blue color of the sky. Rayleigh scatter is also responsible
for haze in images. In aerial photography special filters are used to filer out the scatter blue light
to reduce haze. In digital images there are different techniques used to minimize the impacts of
Rayleigh scatter.

At sunrise and sunset the incoming sunlight travels a longer distance through the atmosphere. The
longer path produces leads to scatter of the short (blue) wavelengths that is so complete we only
see the longer wavelengths of light, the red and orange. In the absence of particles and scattering
the sky would appear black.

Mie scatter occurs when the wavelength of the electromagnetic radiation is a similar size to the
atmospheric particles. Mie scatter generally influences radiation from the near UV through the
mid- infrared parts of the spectrum. Mie scatter mostly occurs in the lower portions of the
atmosphere where larger particles are more abundant, and dominates when cloud conditions are
overcast. Pollen, dust and smog are major cause of mie scatter. Mie scatter produces general haze
in images.

Non-selective scattering occurs when the diameter of the particles in the atmosphere are
much larger than the wavelength of radiation. Non-selective scatter is primarily caused by water
droplets in the atmosphere. Non-selective scatters scatter all visible light evenly - hence the term
non-selective. In the visible wavelength light is scattered evenly, hence fog and clouds appear
white. Since clouds scatter all wavelengths of light this means that clouds block all energy from
reach the Earth's surface. This can make interpreting and analyzing remote sensed imagery difficult
in areas prone to cloud cover. Clouds also cast shadows that change the illumination and relative
reflectance of surface features. This can be a major limitation in remote sensing imagery.

[Marks 30]
(iii). There are three common methods for resampling: nearest neighbor, bilinear interpolation,
and cubic convolution.
Nearest neighbor resampling uses the digital value from the pixel in the original image which is
nearest to the new pixel location in the corrected image. This is the simplest method and does not
alter the original values, but may result in some pixel values being duplicated while others are lost.
This method also tends to result in a disjointed or blocky image appearance.

Bilinear interpolation resampling takes a weighted average of four pixels in the original image
nearest to the new pixel location. The averaging process alters the original pixel values and creates
entirely new digital values in the output image. This may be undesirable if further processing and
analysis, such as classification based on spectral response, is to be done. If this is the case,
resampling may best be done after the classification process.
Cubic convolution resampling goes even further to calculate a distance weighted average of a
block of sixteen pixels from the original image which surround the new output pixel location. As
with bilinear interpolation, this method results in completely new pixel values. However, these
two methods both produce images which have a much sharper appearance and avoid the blocky
appearance of the nearest neighbor method.

[Marks 30]
(iv).
The purpose of georeferencing is to transform the image coordinate system (u,v), which may be
distorted due to the factors discussed above, to a specific map projection (x,y) as shown in
Figure below. There are two kinds of geometric corrections procedures such as Image-to-image
registration and Image-to-map registration.

Image-to-image registration refers to transforming one image coordinate system into another
image coordinating system. Image-to-map registration refers to transformation of one image
coordinate system to a map coordinate system resulted from a particular map projection.

[Marks 20]
Question 3
(i). High Frequency filters in special domain
Spatial filtering is the process of altering of pixel values based upon spatial characteristics for the
purpose of image enhancement.
High Frequency filters designed to emphasize high spatial frequency by emphasizing abrupt local
changes in gray level values between pixels. High Frequency filters preserve high frequencies and
removes slowly varying components and it emphasizes fine details. High Frequency filters used
for edge detection and enhancement. Edges are locations where transition from one category to
other occurs. There are two types of high pass filters linear and nonlinear. In linear filters output
brightness value is a function of linear combination of BV’s located in a particular spatial pattern
around the i,j location in the input image. In nonlinear output brightness value is a function of
nonlinear combination of BV’s. In edge detection brightness will be lost. Delineates Edges and
makes the shapes and details more prominent however.

(ii). Linear Contrast Stretch

Most satellite sensors are designed to accommodate a wide range of illumination conditions, from
dark boreal forest to highly reflective desert regions. Pixel values in most scenes occupy a small
range of values. This results in low display contrast. A contrast enhancement expands the range
of “displayed” pixel values and increases image contrast.
The simplest type of enhancement is a 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 transformation to stretch this range to fill the full range. In the following
example, the minimum value (occupied by actual data) in the histogram is 84 and the maximum
value is 153. These 70 levels occupy less than one-third 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. This
enhances the contrast in the image with light toned areas appearing lighter and dark areas appearing
darker, making visual interpretation much easier.

[Marks 20]
(iii) Edge Enhancement
For many remote sensing Earth science applications, the most valuable information that may be
derived from an image is contained in the edges surrounding various objects of interest. Edge
enhancement delineates these edges. Edges may be enhanced using either linear or nonlinear edge
enhancement techniques.
There are two kinds of linear edge enhancement techniques. They are
- Directional
Vertical Horizontal or any direction
- Laplacian
Highlight points, lines, edges, suppress uniform and smooth regions
(iv). Band Rationing

Sometimes differences in brightness values from identical surface materials are caused by
topographic slope and aspect, shadows, atmospheric constitutional change, or season’s changes in
sun angle and intensity. Band ratio can be applied to reduce the effects of such environmental
conditions. In addition, band ratio also help to discriminate between soils and vegetation

BVi , j ,k
BVi , j ,ratio =
BVi , j ,l

Where:
- BVi,j,k is the original input brightness value in band k
- BVi,j,l is the original input brightness value in band l
- BVi,j,ratio is the ratio output brightness value [Marks 20]

(v). Principal Component Analysis

The multispectral image data is usually strongly correlated from one band to the other. The level
of a given picture element on one band can to some extent be predicted from the level of that same
pixel in another band. Principal component analysis is a pre-processing transformation that creates
new images from the uncorrelated values of different images. This is accomplished by a linear
transformation of variables that corresponds to a rotation and translation of the original coordinate
system.
Principal component analysis operates on all bands together. Thus, it alleviates the difficulty of
selecting appropriate bands associated with the band rationing operation. Principal components
describe the data more efficiently than the original band reflectance values. The first principal
component accounts for a maximum portion of the variance in the data set, often as high as 98%.
Subsequent principal components account for successively smaller portions of the remaining
variance. Principal component transformations are used for spectral pattern recognition as well as
image enhancement. Principal component enhancement techniques are particularly appropriate in
areas where little a priori information concerning the region is available.
 [Marks 20]
Question 4

(i).
Two major categories of image classification techniques include unsupervised (calculated by
software) and supervised (human-guided) classification.

Unsupervised classification is where the outcomes (groupings of pixels with common

characteristics) are based on the software analysis of an image without the user providing sample
classes. The computer uses techniques to determine which pixels are related and groups them into
classes. The user can specify which algorism the software will use and the desired number of
output classes but otherwise does not aid in the classification process. However, the user must have
knowledge of the area being classified when the groupings of pixels with common characteristics
produced by the computer have to be related to actual features on the ground (such as wetlands,
developed areas, coniferous forests, etc.).
Supervised classification is based on the idea that a user can select sample pixels in an image that
are representative of specific classes and then direct the image processing software to use these
training sites as references for the classification of all other pixels in the image. Training sites (also
known as testing sets or input classes) are selected based on the knowledge of the user. The user
also sets the bounds for how similar other pixels must be to group them together. These bounds
are often set based on the spectral characteristics of the training area, plus or minus a certain
increment (often based on "brightness" or strength of reflection in specific spectral bands). The
user also designates the number of classes that the image is classified into. Many analysts use a
combination of supervised and unsupervised classification processes to develop final output
analysis and classified maps.
Steps involved in supervised classification
1. Definition of Information Classes
2. Training/Calibration Site Selection - delineate areas of known identify on the digital image
3. Generation of Statistical Parameters - define the unique spectral characteristics (signatures)
4. Classification - assignment of “unknown” pixels to the appropriate information class
5. Accuracy Assessment - test/validation data for accuracy assessment
6. Output Stage

Steps involved in Unsupervised Classification

1. Definition of minimum and maximum number of categories to be generated by the
particular classification algorithm (based on an analyst’s knowledge or user requirements).
2. Random selection of pixels to form cluster centres.
3. Algorithm then finds distances between pixels and forms initial estimates of cluster centers
as permitted by user defined criteria.
4. As pixels are added to the initial estimates, new class means are calculated. This is an
iterative process until the mean does not change significantly from one iteration to the next
[Marks 40]
(ii)
Numerous mathematical approaches to spectral pattern recognition have been developed for
supervised classification.
1. Minimum distance to mean
2. Parallelepiped classifier
3. Maximum likelihood classifier

Minimum distance to mean classifier

1. Calculate of the mean spectral value in each band and for each category.
2. Relate each mean value by a vector function
3. A pixel of unknown identity is calculated by computing the distance between the value of the
unknown pixel and each of the category means.
4. After computing the distances, the unknown pixel is then assigned to the closest class.
 Limitations of this process include the fact that it is insensitive to different degrees of variance
within spectral measurements.

A “centroid” for each class is determined from

the data by calculating the mean value by
band for each class. For each image pixel, the
distance in n-dimensional distance to each of
these centroids is calculated, and the closest
centroid determines the class.
Benefits: mathematically simple and
computationally efficient
Drawback: insensitive to different degrees of
variance in spectral response data.

Parallelepiped classifier
In this classifier, the range of spectral measurements are taken into account. The range is defined
as the highest and lowest digital numbers assigned to each band from the training data
An unknown pixel is therefore classified according to its location within the class range. However,
difficulties occur when class ranges overlap. This can occur when classes exhibit a high degree of
correlation or covariance.
This can be partially overcome by introducing stepped borders to the class ranges.
The minimum and maximum DNs for each
class are determined and are used as thresholds
for classifying the image.
Benefits: simple to train and use,
computationally fast
Drawbacks: pixels in the gaps between the
parallelepiped cannot be classified; pixels in
the region of overlapping parallelepiped cannot
be classified.
Maximum Likelihood Classifier
This classifier quantitatively evaluates both the variance and covariance of the trained spectral
response patterns when deciding the fate of an unknown pixel.
To do this the classifier assumes that the distribution of points for each cover-type are normally
distributed
Under this assumption, the distribution of a category response can be completely described by the
mean vector and the covariance matrix.
Given these values, the classifier computes the probability that unknown pixels will belong to a
particular category.
Max likelihood uses the variance and covariance in class spectra to determine classification
scheme. It assumes that the spectral responses for a given class are normally distributed.
We can then determine a probability surface, where for a given DN, being a member of a particular
class. The pixel is classified by using the most likely class or “Other” if the probability isn’t over
some threshold.
Benefits: takes variation in spectral response into consideration
Drawbacks: computationally inefficient, multimodal or non-normally distributed classes can be
misclassified.

Question 5
(i). A Radar system has three primary functions:

- It transmits microwave (radio) signals towards a scene

- It receives the portion of the transmitted energy backscattered from the scene
 - It observes the strength (detection) and the time delay (ranging) of the return signals.

[Marks 20]

(ii).

Principles of Real Aperture Radar

The two main types of radar images are the circularly scanning plan-position indicator (PPI)
images and the side-looking images. The PPI applications are limited to the monitoring of air and
naval traffic. Remote sensing applications use the side-looking images which are divided into two
types--real aperture radar (usually called SLAR for side-looking airborne radar or SLR for side-
looking radar) and synthetic aperture radar (SAR).

The SLAR imaging system uses a long, straight antenna mounted

on a platform (aircraft or satellite), with its longitudinal axis
parallel to the flight path. The antenna emits pulses of
electromagnetic energy directed perpendicular to the flight path
of the platform and downward to the surface of the Earth, as
illustrated in Figure. These pulses fall within a narrow area on the
ground and are scattered, usually in many directions, including
the direction of the antenna. The return echoes arrive at the
antenna at different times, depending on the distance from the
antenna to the specific scattering object on the ground.

Principles of Synthetic Aperture Radar

SAR takes advantage of the Doppler history of the radar

echoes generated by the forward motion of the spacecraft to
synthesise a large antenna (see Figure). This allows high
azimuth resolution in the resulting image despite a
physically small antenna. As the radar moves, a pulse is
transmitted at each position. The return echoes pass through
the receiver and are recorded in an echo store.

SAR requires a complex integrated array of onboard

navigational and control systems, with location accuracy provided by both Doppler and inertial
navigation equipment
Compared to RAR, SAR synthetically increases the antenna's size to increase the azimuth
resolution though the same pulse compression technique as adopted for range direction. Synthetic
aperture processing is a complicated data processing of received signals and phases from moving
targets with a small antenna. The resulting azimuth resolution is given by half of real aperture radar
as shown as follows:

- Real beam width β =λ/D

- Real resolution ΔL = β . R = Ls (synthetic aperture length)

- Synthetic beam width βs = λ / 2 . Ls = D / (2 . R)

- Synthetic resolution Δ Ls = βs . R = D / 2

where λ is the wavelength, D the radar aperture, and R the distance antenna-object.

This is the reason why SAR has a high azimuth resolution with a small size of antenna regardless
of the slant range, or very high altitude of a satellite.

Sideways Looking Airborne Radar ( SLAR) mmeasures range to scattering targets on the ground,
can be used to form a low resolution image whereas Synthetic Aperture Radar has same principle as
SLAR, but uses image processing to create high resolution images.

(iii) The spatial resolution of RAR is primarily determined by the size of the antenna used: the
larger the antenna, the better the spatial resolution. Other determining factors include the pulse
duration (τ) and the antenna beam width.

Range Resolution
Range resolution is defined as

c pulse length  speed of light

Rr = =
2 cos  2 cos ( depression angle )

Where c is the speed of light, and γ the

incidence angle.
Azimuth resolution

Azimuth resolution is defined as

S 
Ra =
L
slant range  wavelength
antenna length
where L is the antenna length, R the distance antenna-
object, and λ is the wavelength. For systems where the
antenna beam width is controlled by the physical
length of the antenna, typical resolutions are in the
order of several kilometres.

[Marks 20]

(iv).
Advantages of radar
• All weather, day or night
– Some areas of Earth are persistently cloud covered
• Penetrates clouds, vegetation, dry soil, dry snow
• Sensitive to water content, surface roughness
– Can measure waves in water
• Sensitive to polarization and frequency
• Interferometry (later) using 2 receiving antennas
Disadvantages of radar
• Penetrates clouds, vegetation, dry soil, dry snow
– Signal is integrated over a depth range and a variety of materials
• Sensitive to water content, surface roughness
– Small amounts of water affect signal
 – Hard to separate the volume response from the surface response
• Sensitive to polarization and frequency
– Many choices for instrument, expensive to cover range of possibilities
– The math can be formidable
[Marks 20]
(v).
Foreshortening
• In flat terrain, easy to convert a slant-range radar image into a ground-range radar image
– … but with trees, tall buildings, or mountains, you get radar relief displacement
– the higher the object, the closer it is to the radar antenna, and therefore the sooner
(in time) it is detected on the radar image
• Terrain that slopes toward the radar will appear compressed
or foreshortened compared to slopes away from the radar

Foreshortening factor Ff = sin ( −  )

where  = slope angle, + toward radar, − away

Layover

• Extreme case of foreshortening, when incidence angle is less

than slope angle toward radar (i.e. θ<α)
– cannot be corrected
– got to be careful in the mountains

Shadow

• When slope away from radar is steeper than the depression

angle, i.e. –α > γ

[Marks 20]