Scribd
Upload
21 views29 pages

Week-3 Module-13 Geo-referencing Technique

The document discusses georeferencing techniques for transforming satellite images from geometric to geographic coordinate systems, addressing both systematic and nonsystematic geometric errors. It outlines the process of image registration using ground control points and various resampling methods, including Nearest Neighbour, Bi-linear, and Cubic Convolution, each with its own advantages and drawbacks. Additionally, it emphasizes the importance of root mean square error in assessing transformation accuracy and the need for quality control points.

Uploaded by

password123reset
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
21 views29 pages

Week-3 Module-13 Geo-referencing Technique

The document discusses georeferencing techniques for transforming satellite images from geometric to geographic coordinate systems, addressing both systematic and nonsystematic geometric errors. It outlines the process of image registration using ground control points and various resampling methods, including Nearest Neighbour, Bi-linear, and Cubic Convolution, each with its own advantages and drawbacks. Additionally, it emphasizes the importance of root mean square error in assessing transformation accuracy and the need for quality control points.

Uploaded by

password123reset
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 29

Remote Sensing and Digital Image Processing of Satellite Data

Geo-referencing Technique

Dr. Arun K. Saraf,


Professor
Department of Earth Sciences

1
GEOREFERENCING
Georeferencing transforms images / maps from
geometric coordinate system to geographic
coordinate system using base map / image
(having geographic coordinates).
When an image is acquired from satellite, the
resulting image has certain systematic and
nonsystematic geometric errors introduced
through sensor distortion, scan skewness,
panoramic distortion and attitude of the platform
(velocity, altitude, pitch, roll and yaw).
SYSTEMATIC DISTORTIONS
NON-SYSTEMATIC DISTORTIONS
• Images are stored as raster data, where each pixel in the
image has a row and column number and hence are in
geometric coordinate system.

• In order to display and analyse images with other


georeferenced maps / datasets, it is necessary to
establish an image-to-world transformation that
converts the image coordinates to real-world
coordinates.
• A common method of image georeferencing / geometic
correction / image registration / rectification is to
statistically find a polynomial of a given order that
minimizes the error in a transformation from the original
image coordinates to the rectified image coordinates.

• The transformation is found by performing a least


squares fit for the coefficients of the given polynomial
using ground control points (GCPs) that are picked by the
user.
RESAMPLING
Original data

First order Second order Third order


Polynomial Polynomial polynomial
(conformal) (affine)
10
Number of Ground Control Points (GCPs) required
= [(P+1)(P+2)] / 2
Model Order No. of GCPs
required
1 3
2 6
3 10
4 15
5 21
• Once the transformation is found, it is applied for
every pixel in the input image.
• The other operation to perform when doing a
transformation of this type is determining the pixel
value.
• This is accomplished through using resampling
techniques (e.g. nearest neighbour, bilinear or cubic
convolution).
The transformation can be represented by a polynomial
of order m such as:
m m j
x    a jk x y
, j k

j 0 k 0
m m j
y    b jk x y
, j k

j 0 k 0
For example: In ArcGIS the image-to-world
transformation is a six-parameter Affine
Transformation (second order polynomial
equation) in the form of:
x1 = Ax +By + C
y1 = Dx + Ey + F
where,
X1 = calculated x-coordinate of the pixel on the map
y1 = calculated y-coordinate of the pixel on the map
x = column number of a pixel in the image
y = row number of a pixel in the image
A = x-scale; dimension of a pixel in map unites in x direction
B, D = rotation terms
C,F = translation terms; x, y map coordinates of the center of the upper-left pixel
E = negative of y-scale; dimension of a pixel in map units in x direction
The y-scale (E) is negative, because the origins of
an image and a geographic coordinate system are
different. The origin of an image is located in the
upper-left corner, whereas the origin of the map
coordinate system is located in the lower-left
corner.
Interpreting the root mean square error
• When the general formula is derived and applied to the control
point, a measure of the error—the residual error—is returned.

• The error is the difference between where the from point


ended up as opposed to the actual location that was
specified—the to point position.

• The total error is computed by taking the root mean square


(RMS) sum of all the residuals to compute the RMS error.
• This value describes how consistent the transformation
is between the different control points (links).
• When the error is particularly large, you can remove
and add control points to adjust the error.
• Although the RMS error is a good assessment of the
transformation's accuracy, don't confuse a low RMS
error with an accurate registration, e.g. the
transformation may still contain significant errors due
to a poorly entered control point.
The more control points of equal quality
used, the more accurately the polynomial
can convert the input data to output
coordinates.
RESAMPLING METHODS
RESAMPLING METHODS
Nearest Neighbour

Nearest Neighbour resampling


determines the pixel value from
the closest pixel to the input
coordinate specified, and assigns
that value to the output
coordinate.
• This method is considered the most efficient
procedure in terms of computation time.

• Nearest Neighbour does not alter the pixel value.


• This is desirable if subtle changes in pixel values need to be
retained.

• This method, however, induces a small error into the


corrected image.

• The corrected image may be offset spatially by up to half


pixel.

• The corrected image may be jagged or blocky in appearance


if there is much rotation and / or scale change.
Bi-linear resampling
• Bi-linear resampling determines a
weighted average of the fours
nearest pixels in the uncorrected
image.
• Closer the central points of pixels,
the greater contribution or weight it
will have to the final DN value to be
assigned to the corrected pixel.
• Bi-linear resampling generates a smoother-appearing
resampled image, its pixel value is altered in the process,
resulting in a blurring or loss of image resolution.

• This method requires three to four times the computation


time as compare to Nearest Neighbour method.

• Highly accurate registration will achieve more faithful pixel


values from the original uncorrected image.
Cubic Convolution

This more sophisticated method


uses the weighted average of the
sixteen surrounding pixels of the
uncorrected image to approximate
the pixel value of the new pixel
space in the corrected image.
• It is closer to the perfect sin(x) / x resampler than the
Nearest Neighbour or Bi-linear resampling and avoids
the disjointed appearance of the Nearest Neighbour
method.

• It provides a slightly sharper image than the bilinear


method but it also corrupts the original pixel values.
• This method is not recommended if classification is
to follow as the new pixel values may be slightly
different from the actual radiance values detected
by the satellite sensor.

• The computation time of this procedure is about ten


times greater than for the Nearest Neighbour
method.
THANKS

You might also like