You are on page 1of 94

Spatial Resolution

Radiometric resolution

Multispectral Vs Hyperspectral Image

Concepts of Aerial Photography

Scale of photograph :

Fiducial Marks
small registration marks exposed on the edges of a photograph

Flight plan

Overlap in flight plan

Aerial Photograph mosaic

Stereoscopy
The ability to see and appreciate depth of field through the
perception of parallax.

Photogrammetry using Stereo pairs

Modern Stereoscopy

Visual Image Interpretation


act of examining photographic images for the purpose of identifying
objects and judging their significance.
Elements of visual image interpretation :

location
size
Shape
shadow
tone/color
Texture
pattern
height/depth
site/situation/association

Size
Size of objects in an image is a function of scale.
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

Shape
Shape refers to the general form,
structure, or outline of individual
objects.
Shape can be a very distinctive clue
for interpretation.

Tone/Color
Tone refers to the relative
brightness or colour of objects in
an image.
Variations in tone also allows the
elements of shape, texture, and
pattern of objects to be
distinguished.

Tone vs Color

Texture
Texture refers to the arrangement and
frequency of tonal variation in
particular areas of an image.
Rough textures would consist of a
mottled tone where the grey levels
change abruptly in a small area.
Smooth textures would have very
little tonal variation.

Pattern
Pattern refers to the spatial
arrangement of visibly
discernible objects.
Typically an orderly repetition of
similar tones and textures will
produce a distinctive and
ultimately recognizable pattern

Shadow
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.
Shadows can also reduce or eliminate
interpretation in their area of
influence

Site /Situation/Association
Association takes into account
the relationship between other
recognizable objects or features
in proximity to the target of
interest

Visual Image interpretation


Advantages
Interpreters knowledge are available
Excellent in spatial information extraction
Limitations
Time consuming
Individual difference
Simultaneous analysis of multiple bands/images is difficult
Serious biased error may introduce due to analysts own assumption or
expectation

Digital Image interpretation


Digital image processing is the application of algorithms on digital
images to perform processing, analysis, and information extraction.
Data must be recorded and available in digital form.
Data recorded on a photographic film can also be converted into
digital form, but a very few of the digital processing techniques can be
applied on them.

Advantages of Digital Image Processing


Short processing time

Re-productivity
Extraction of physical quantities
Data commonly transmitted or converted to digital format

Analysis of individual points (pixel)


Analysis of multiple bands/images in a single platform
Handling of large data volumes

Accuracy assessments

Digital Image

Pixels
The smallest two-dimensional
non-divisible element of an
image is called pixel.
Each pixel stores a digital
number (DN) measured by the
sensor.
Represents individual areas
scanned by the sensor.
Smaller pixel size offers greater
spatial accuracy.

Process of Digital Image Processing

Pre - processing
In their raw form, remotely sensed data generally contain flaws or
deficiencies. The correction of deficiencies and the removal of flaws
present in the data are termed pre-processing.
Pre-processing includes:
Radiometric corrections
Geometric corrections
Miscellaneous pre-processing

Radiometric Correction
The main purpose for applying radiometric corrections is to reduce
the influence of errors or inconsistencies in image brightness values.
Radiometric errors and inconsistencies are often referred to as
noise.
Noise means any undesirable variation in image brightness.

De striping
Striping occurs if a detector goes
out of adjustment.
Individual detectors appear
lighter or darker than their
neighboring detectors
Correction is applied by selecting
one sensor as a standard and
adjusting the brightness for all
other pixels.
Independent reference value of
brightness can also be used.

Removal of Missing Scan Lines


Missing scan line occurs when a
detector either completely fails
to function, or becomes
temporarily saturated during a
scan.
Corrected by replacing the bad
line with a line of estimated data
file values, which is based on the
lines above and below it.

Random Noise Removal


Odd pixels that have varying DNs
frequently in images, and if they
are not systematic, they can be
considered as random noise.
Noisy pixels can be replaced by
substituting an average value of
neighboring DNs.

Atmospheric Correction
Atmospheric effects are not
considered as errors.
They are part of the signal
received by the sensing device.
A number of algorithms have
been developed to correct
atmospheric effects.

Before

Courtesy: CCRS

After

Examples of atmospheric correction

Geometric Correction
Digital images often contain systematic and non-systematic geometric
errors that arise from the earth curvature, platform motion, relief
displacement, non-linearities in scanning motion, the earth rotation, etc.
Digital images are not geographically referenced.
Removing these errors is known as geometric correction.

Systematic Correction
Systematic errors:

Scan skew
Known mirror velocity variation
Earth-rotation skew
Platform velocity variation

Systematic distortions are corrected by applying formulas derived by


modelling the sources of distortions mathematically.

Non Systematic Errors


Establishing the relationship between two different coordinate
systems.
Two approaches:

Image to ground geo-correction (georeferencing)

Image to image correction (registration)

Image to Ground Geo-correction


Correction is the process of giving an image a Real World
coordinate system

116 41 52.20 W, 33 50 03.23 N


(Longitude, Latitude)

Image to Image Correction


Assigning the co-ordinate system of one image to a second image of
the same area

Fitting the co-ordinate


system to another image

Subsetting
Breaking out a portion of a
large file/image into one or
more smaller files/images.

Area of interest

Mosaicking
Combining multiple scenes to cover a larger area.

Image Enhancement
Image enhancement can be defined as the conversion of the image
quality to a better and more understandable level for feature
extraction or image interpretation.
Enhancement is generally applied to single-band images or separately
to individual bands
Principle objective is to process an image so that the result is more
suitable than the original image for a specific application.

Procedures
Two type of procedures :
Point operations
Local operations
Point operations : When we change the value of each individual pixel
independent of all other pixels.
Local Operations : When we change the value of individual pixels in the context
of the values of neighboring pixels.

Image Reduction
Original image
30 55 35 76 48 89 98 36 33 76

2x image reduction

87 34 55 98 45 75 62 98 78 12
12 69 87 36 87 69 47 69 78 98

Reduced image

39 87 95 22 36 14 65 68 38 35

30 35 48 98 33

54 92 80 60 34 43 78 53 58 50

12 87 87 47 78

69 97 65 98 32 72 91 38 78 65

54 80 34 78 58

66 54 57 85 95 36 96 16 3

66 57 95 96 3

5 75 98 75 98 78 89 65 88 9

Image Magnification
2x image magnification
Original image

Magnified image

30 35 48 98 33

30 30 35 35 48 48 98 98 33 33

12 87 87 47 78

30 30 35 35 48 48 98 98 33 33

54 80 34 78 58

12 12 87 87 87 87 47 47 78 78

66 57 95 96 3

12 12 87 87 87 87 47 47 78 78
54 54 80 80 34 34 78 78 58 58
54 54 80 80 34 34 78 78 58 58
66 66 57 57 95 95 96 96 3

66 66 57 57 95 95 96 96 3

Color Compositing
A color image can be generated by compositing three selected bands
of multi-band image and with the use of three primary colors (RGB).
Display contains three color guns : RGB , hence only 3 bands can be
seen at a time.
Color Composites :
Additive Color composites : which use RGB
Subtractive Color composites : use three pigments of RGB : cyan , magenta
and yellow

True Color Image vs False color image


True Color : A combination where image captured in blue band is passed though
blue gun , red from red gun and green from green gun.
False color : when one of the primary color is not present , mathematical
combinations are made. Eg if blue is not present :
RED = Red band
GREEN = 0.75 Green band + 0.25 NIR band
BLUE = 0.75 Green band 0.25 NIR band
Infrared Color Composite : BLUE = Green Band
GREEN = Red Band
RED = Infrared Band

True color vs False color

Infrared color composite

Transect Extraction
Pixels that lie on a transect can be
measured and displayed to compare
spectral or spatial differences.

Band-1

Band-2

Band-4

Band-3

Contrast Enhancement
The range of brightness values present on an image is referred to as
contrast.
Contrast enhancement is a process that makes the image features
stand out more clearly by making optimum use of the color
intensities available on the display or output device.
Look-up Table (LUT) computer stores new values are stored in LUT
and utilizes these values to display the image.

Histogram
A histogram is a graph of data frequency or distribution.
A statistical graphic representation of the range of tones from dark to
light and associated number of pixels for each tone for an image.

Image contrast and Histogram


Contrast: The range of brightness values present on an image

Image Contrast

Contrast Enhancement

Contrast manipulations involve changing the range of values in an image in order


to increase the contrast.

Contrast Enhancement

Types of contrast enhancement


Linear contrast enhancement

Minimum-maximum linear contrast stretch

Percentage linear contrast stretch

Average and standard deviation stretch

Piecewise linear contrast stretch

Nonlinear contrast enhancement

Histogram equalization

Histogram normalization

Reference stretch

Density slicing

Thresholding

Minimum Maximum Stretch

Saturation Stretch

Average and Standard Deviation Stretch

Piecewise Stretch

Different Contrast Enhancements

No stretch

20% tail trim

Min-max stretch

1 x standard deviation

5% tail trim

2 x standard deviation

Filtering
Filtering is a process of changing the spatial frequency.
Purposes:

To improve interpretability of image data

To highlight or suppress specific features of an image based on


spatial frequency
Methods:

Convolution filtering

High Pass

Low Pass

Edge Detection

their

Spatial Frequency

Zero spatial frequency

Low spatial frequency

High spatial frequency

Convolution Filtering
It means moving a window of set of pixels in dimensions (3x3 , 5x5)
over each pixel in the image, applying a mathematical calculation
using the pixel values under that window, and replacing the central
value with that pixel.
This window is known as convolution kernel

Low Pass filtering


Low-frequency kernels are applied to decrease the spatial frequency
(provides smooth image).

Low Pass Filtering

Before filtering

After filtering

High Pass Filter


It is the opposite of low pass filter ( sharpens the image)
High-frequency kernels are applied to increase the spatial frequency
(provides sharper image).

High Pass Filtering

Before filtering

After filtering

Edge Detection Filtering


They are used to highlight linear features like roads, field
boundaries,etc.
Zero Sum kernels are used and division is not applied.
This generally causes the output values to be zero in areas where all
input values are equal, low in areas of low spatial frequency and
extreme in areas of high spatial frequency.

Edge Detection

Before filtering

After filtering

Image Transformation
Image transformations generate new images from two or more sources
which highlight particular features or properties of interest, better than the
original input images.

Common transformations:

Image arithmetic operations

Principal component transformation (PCT)

Tasselled cap transformation (TCT)

Colour space transformation (CST)

Fourier transformation

Image fusion

Arithmetic Operations
Addition
Subtraction
Multiplication
Division
The images may be separate spectral bands from a single multispectral data set or they may be individual bands from image data
sets that have been collected at different dates.

Image Addition (Averaging)


Addition is getting new DN value of a pixel in the output image by
averaging the DN values of corresponding pixels of input images

Image Subtraction (Change Detection)


It is the process of subtracting the DN value of one image from
corresponding pixel of other image.

Change Detection
1987 1987

1997

Change

Image Multiplication
In this process pixel-by-pixel multiplication of two images is
performed.

Indices and Rationing


Used to create output image by mathematically combining the DN
values of different bands.
Eg. Vegetation index DNNIR / DNR

Normalized differential vegetation index (DNNIR DNR) / (DNNIR + DNR)

Concept of Indices

Image Classification
Image classification is sorting pixels into a finite number of individual
classes, or categories of data, based on their DN values.

Continuous image

Thematic image

Supervised vs Unsupervised Classification


Supervised
Have a set of desired classes in mind and then create the appropriate
signatures from the data.
when one wants to identify relatively few classes
when one has selected training sites that can be verified with ground truth
data
when one can identify distinct, homogeneous regions that represent each
class.

Unsupervised
Classes to be determined by spectral distinctions that are inherent in the data
and define the classes later.
when one wants to define many classes easily, and then identify classes.

Training for Classification


Computer system must be trained to recognize patterns in image data.
Process of defining the criteria by which these patterns are recognized.

Supervised Training is controlled by the analyst.


Select pixels that represent patterns and instruct the computer system to
identify pixels with similar characteristics.
More accurate but requires high skill.

Unsupervised Training is computer-automated.


Specify number of classes and the computer uncovers statistical classes.
Less accurate and less skill required.

Supervised Classification
Raw
Data

Prepro-

Signature

-cessing

Collection

Classification

Signature
Evaluation

Signature
A set of pixels selected to represent each primary land-cover.

Lake

Knowledge

Field Data

Personal Experience
Photos

Previous Studies

Agriculture
River
Forest

Minimum Distance Classifier

Unsupervised Classification
Clustering algorithms are used in unsupervised classification.
Spectral values of pixels are grouped first and then matched by the
analyst to any specified class.

Analyst specifies how many classes are required and threshold values
of variation within the clusters and among the clusters.
If a cluster needs to be broken , or clusters need to be combined
analyst can take the decision.

Accuracy Assessment
Accuracy assessment of remote sensing product is a feedback
system for checking and evaluating the objectives and the results.

Contingency Matrix
Contingency matrix for accuracy assessment