Image Restoration

Question 1

Errors related to the sensor/platform (3 Sources of Error in Satellite Imagery)

Answer 1

Errors related to the sensor/platform:
–Variations in orbit altitude, velocity and orientation -> changes in scale or pixel’s positions. Hard to correct.
– Calibration issues: stripping effects or missed pixels

Question 2

Earth’s rotation and curvature (3 Sources of Error in Satellite Imagery)

Answer 2

Earth’s rotation and curvature:
– Earth’s curvature can cause changes in the size of
pixels within the same image (the Earth does not stop
rotating while the image is being acquired!)

Question 3

Atmospheric effects (3 Sources of Error in Satellite Imagery)

Answer 3

Atmospheric effects:
– Radiance can be scattered or absorbed in the atmosphere affecting the energy reaching the Earth’s surface as well as the sensor.
– Results: attenuation of the signal, reduced contrast in the image, haze, etc

Question 4

Image Preprocessing

Answer 4

The problem:
Raw satellite imagery is subject to geometric distortions, radiometric inconsistencies, clouds, and other noise.

• Preprocessing = Operations performed prior to analysis
Objective: correct for sensor and platform - specific distortions, specifically:

– Geometric distortions: spatial fidelity
– Radiometric distortions: fidelity of EM energy measurements (e.g., atmospheric effects)

Question 5

Geometric corrections

Answer 5

• Necessary to remove geometric distortion due to variations in the sensor - Earth geometry (e.g., change of altitude) or incorrect conversion of data to Earth’s surface coordinates.
• Objective : ensure that the pixels are in their proper
  geographical locations
  – Allows RS derived information to berelated to other spatial information (e.g., other images or GIS layers)
  – Geometrically corrected imagery can be used to extract accurate distance, polygon area, and direction information.

Several Geometric correction methods.
Our focus: image registration (also known as geometric correction by resampling
)

Question 6

Image Registration

Answer 6

• Involves matching the points in the image to be corrected to known (or “true”) ground coordinates.
• Requires: Pairs of resulting of matching points= Ground Control Points or GCPs.
• The coordinate system of the image to be corrected will be transformed based on the relationship between the coordinates of the uncorrected image and the ‘true” coordinates (GCPs).
• Requires a source of “true” coordinates such as paper or digital maps, GPS points or another image (problem: Geometric errors of the reference image will be inherited as well!)

Geometric or image Registration25
2 parts:
1) generate a new (empty) grid with correct coordinates
2) transpose the original values from the uncorrected image to the created (empty) grid (resampling)

Question 7

neighbor resampling vs. bilinear interpolation

Answer 7

Nearest neighbor approach:
• Advantages:
– Original data are retained, recommended
before classification.
– Easy to compute and therefore fast

• Disadvantages:
– Produces an image with rough appearance
relative to the original
– Data values may be lost or duplicated,
which may result in breaks in linear features such as
roads, streams, and boundaries

Advantages:
•Stair
- step effect caused by the nearest neighbor approach is reduced.
• Image looks smooth.

Disadvantages:
• Alters original data and reduces contrast by averaging neighboring values together.
• Is computationally more intensive than nearest neighbor.

Question 8

Radiometric correction

Answer 8

Objective: correct for sensor and platform
- specific distortions, specifically:
– Geometric distortions: spatial fidelity
– Radiometric distortions: fidelity of EM energy measurements (e.g., atmospheric effects) 40

• Radiometric distortion:
– Is specific to the conditions during data acquisition
(e.g., illumination and viewing geometry) and – Sensor used (e.g, sensor noise and calibration issues)