Image Corrections/Errors

Question 1

Radiometric issues: 4

Answer 1

instrument noise - sensitivity error
Striping
Dropline
Bit Errors

Question 2

Atmospheric (and other radiation-related) disturbances: 4

Answer 2

‣accommodation for atmospheric noise/disturbance
‣accommodation for bidirectional reflectance
‣slope correction
‣accommodation for path radiance

Question 3

How to correct instrument background noise?

Answer 3

Record image with shutter closed. Subtract the recorded values (the noise) from the image.

Question 4

Drop-line correction

Answer 4

using the average based on the line immediately above and below the missing line.

Question 5

Random Bit Errors

How can it be fixed?

Answer 5

These are errors where individual pixels are periodically missing from the image.

How to fix: These values can be replaced by either using an average of the surrounding 9 or 25 pixels (that is using a 3x3 or 5x5 pixel kernel), or by employing a median filter.

Question 6

3 different ways to convert radiance to reflectance?

Answer 6

1. use a radiative transfer model (physics-based) to account for atmospheric constituents,
2. use an empirical (statistical) approach to relate the aircraft measured data to ground measured data,
3. mechanistic approaches.

Question 7

How to convert radiance to reflectance using Empirical Approach?

Answer 7

Choose a reference reflector that is spectrally invariant, that do not change over time, examine differences and subtract difference.

Question 8

Mechanistic Approaches–Path radiance

Answer 8

Subtract ranging values from bands (use histogram to find the values that are far off)

Question 9

Geomatic Corrections: rectification:

Answer 9

the linking, of one “slave” image to another “master” image/map;

Question 10

Geomatic Corrections: Resampling

Answer 10

the transfer of the data from the slave image to the newlycreated, geometrically-correct image.

Question 11

Rectification - 2 steps

Answer 11

• collection of ground control points (GCP)

* generation of a new projection

Question 12

In terms of GCPs

•The degree of the scatter around the best fit regression is expressed by the ______ Error

Answer 12

RMS Error

Question 13

Image Rectification: Conformal Transform
When is it best used?
# of GCPs required?

Answer 13

This transformation is used to translate from one coordinate system to another.

It preserves the shape of the image as well as parallel lines

Best used: approach is used for scale changes, rotation of images and translating to 3-D images.

2 GCPs required.

Question 14

Image Rectification: First Order or Affine Transform:
# of GCPs required?

Answer 14

This transformation is similar to the conformal except that the image can be transformed independently in the X-and Y-directions. The shape of the image is therefore not always preserved.
Only 3 GCPs required. Only three coordinates are required but more can be used to better define the geometric properties of the image.

Question 15

Image Rectification:  Higher Order Polynomial
# of GCPs required?

When is it useful?

Answer 15

• approach does not preserve parallel lines, nor are shapes preserved.
• requires six or more points, depending on the degree of the polynomial.
• useful especially in cases where the geometry differs throughout the image

Question 16

What Happens when you increase Order of Polynomial

Answer 16

By increasing the order of the polynomial, the extent to which one can rubber sheet increases. In other words, the greater the degree of displacement, the higher the order of the polynomial needed to correct it. One must, however, be extremely careful with increasing the degree of the polynomial as the edges of the image become increasing unstable.

Question 17

Image rectification: Orthorectification

Answer 17

A process of rectification which removes the effects of relief/height displacement and lens distortion.

• Lens distortion is removed through the application of correction factors derived though a lens calibration process.
• Relief/height displacement is removed through the use of a digital elevation model or a digital surface model.

Question 18

Image Resampling: Nearest Neighbour

Answer 18

Selects the pixel nearest to the the original one.

Question 19

Image Resampling: Bilinear interpolation

Answer 19

the four pixels surrounding the location of the new pixel are averaged for the new pixel value.

Question 20

Image Resampling: Cubic convolution

Answer 20

• is an improvement on the bilinear interpolation in that a larger number of pixels (16) are used to generate the new pixel value.
• The advantage of this technique is that the image appears to be much less blurred than that obtained with the bilinear interpolation.

Question 21

Choosing Resampling Methods: Which is best used for
A) Nearest Neighbour
B) Bilinear and Cubic Convolution

Answer 21

A) Nearest Neighbour: used when we need to have data that are to be used for modelling and direct measurements
B) Bilinear and Cubic Convolution: used for smoother images - the higher the degree of averaging the greater the smoothing.
• Used when we want to visualize data, want a smoother looking image