Image Transformation
MUHAMMAD HASAN ALI BAIG, D.SC. (PHD)
ASSISTANT PROFESSOR
IGEO
Intro
We are going to transform (change) images according to
our requirement
There are various techniques of transformation
Arithmatic operations
Image addition
Image subtraction
Image multiplication
Image division
Empirical based image transformation
Principal Component Analysis
Hue Saturation Intensity
Image Fusion
Change Detection
Topic : Change Detection
Principal components Analysis (PCA)
•A statistical procedure of data compression.
The first principal component usually accounts for most of the
variability in the data.
•Allow you to pick out patterns and relationships in the
variables.
•Allows you to reduce the side of your dataset
with no significant information loss.
•Isolates and defines features more thoroughly than
conventional Band combinations
Topic : Change Detection
PCA
•A multivariate technique
•Reduces the number of spectral components to
less number of principal components (having the
largest variance) in the input multispectral
images.
•Covariance or correlation matrix is used to
transfer data to an uncorrelated set.
Topic : Change Detection
PCA
•Sort the resulting matrices consisting
of eight vectors in descending order.
•Most of the data variation is expressed
by the first principal component.
Topic : Change Detection
PCA
•The next largest amount of variation is defined by the succeeding
principal Component and this is orthogonal/independent to the
principal component
That precedes it.
•Two approaches are there
1. Separate rotation
2. Merged rotation
Topic : Change Detection
Vegetation Index Differencing
•In this method, ratios called as vegetation indices are computed to
enhance the spectral differences based on strong absorption and
reflection of electromagnetic spectrum.
Many vegetation indices are used, such as:
1. Ratio vegetation index
2. Normalization vegetation index
Vegetation Index Differencing
• This method offers the advantage of emphasizing
The spectral response of different features and
Reduction of impacts of topographic effects and
illumination.
• This method provided very accurate quantitative
Results for delineated forest canopy change but
It was less accurate for delineating forest clear cuts.
Topic : Change Detection
Topic : Change Detection
Topic : Change Detection
Image addition
g(x; y) = f1(x; y) + f2(x; y)
Image subtraction
Gf(i,j) = (255 + G1(i,j) – G2(i,j) ) /2
Image multiplication
If pixel value is less then 127 multiply by zero .
If pixel value is grater then 127 multiply by one.
Image division
This is most common and mostly use
NDVI = (NIR-Red)/(NIR+Red)
Principal Component Analysis
Adjacent bands in multi or hyperspectral remotely sensed image are
generally correlated. Multiband visible/NIR image of vegetated areas
exhibit negative correlations between the NIR and visible red bands
and positive correlations among visible bands because the spectral
reflectance characteristics of vegetation are such that as the vigour or
greenness of vegetation increases the red reflectance diminishes and
the NIR reflectance increases.
Principal Component Analysis
Principal Component Analysis
Principal Component Analysis
Principal Component Analysis
A plot using the axes AB and CD rather than conventional X and Y axes might, in
some cases, prove more revealing of the structures that are present within a data.
Furthermore, if the variation in the direction CD in above Figure contains only a
small portion of the total variability in the data then it may be ignored without
too much lost of information, resulting in data compression.
In above second Figure the number of variables is two but the dimensionality of
the data as shown by the scatter of points is one
Principal Component Analysis
A principle components transform of a multispectral
image might therefore be expected to perform the
following operations:
Estimate the dimensionality of the data set and
Identify the principal axes of variability within the
data.
These Properties of PCA (sometimes also known as
the Karhunen-Loeve transform) mifgt prove to be
useful if the data set is to be compressed.
Principal Component Analysis
The data compression property is useful if more than three spectral bands are
available. A conventional RGB color display system relates a spectral band to one
of the three color inputs (RGB)
Principal Component Analysis
The Landsat TM provides seven bands of data, hence a decision must be made
regarding which three of these seven bands are to be displayed as a colour
composite image.
If the basic dimensionality of the TM data is only three then most of the
information in the seven bands will be expressible in terms of three principle
components. The principle component images could therefore be used to
generate a RGB false color composite.
Principal Component Analysis
Principal Component Analysis
Tasseled Cap Transformation
26
2-D scatter plot was plotted between Red and NIR bands on board.
Different stages of Vegetation growth were discussed starting from the
soil line and ending on soil line.
5/16/2017
Any Questions ??????
Thanks
