Sensing) – Final correctly answered to
pass
Why do we need geometric corrections? - correct answer ✔✔- Allowing overlay with data from
other sources (images are not in a map projection)
- Allowing temporal studies
- Comparison between images
- Creation of basemaps for cartography
Why do we need geometric corrections continued? - correct answer ✔✔- Extraction of position,
direction, or area information
- Spatial analysis
- Identification of unique image features relative to known points
How do we correct for geometric distortions? - correct answer ✔✔By matching coordinates of
features on image and map (GCP's) or using GPS coordinates
Geometric Corrections: Registration? - correct answer ✔✔- Image to image
- Image to map
What are reasons for rectify? - correct answer ✔✔- Comparison between images
- Integration into GIS
- Creation of basemaps for cartography
- Extraction of position, direction, or area information
- Spatial analysis
,- Identification of unique features relative to known points
What is the 3 steps involved in correcting for geometric distortions? - correct answer ✔✔1)
Collect GCP's - used to match image coordinates with map/other image coordinates
2) Rectification: the computation of a new image coordinate matrix, aligned with either an
existing image or some geographic projection (also: warping, rubber-sheeting)
3) Resampling: the population of the new matrix with the brightness values from the original
image file
What are the 3 resampling methods? - correct answer ✔✔1) Nearest Neighbour (NN) - input
value closest to pixel
2) Bilinear Interpolation (BL) - pixel value = weighted average of 4 closest input pixels
3) Bicubic Convolution (CC) - pixel value = weighted calculation based on 16 closest input cells (a
4x4 array)
Nearest Neighbour (NN) - correct answer ✔✔Assigns a pixel value based on nearest pixel in
original matrix to a given pixel in the output matrix (no calculation involved)
What are the advantages of nearest neighbour? - correct answer ✔✔- Simple, and
computationally efficient
- Preserves original input values
- Good for classification
What are the disadvantages of nearest neighbour? - correct answer ✔✔- Image is blocky
- Values may be lost, others may be duplicated
- May result in breaks in linear features (roads, streams, boundaries)
Bilinear Interpolation (BL) - correct answer ✔✔Assigns a value based on a weighted linear
average of the 4 nearest pixels in the original matrix
, What are the advantages of bilinear interpolation? - correct answer ✔✔Smoother image than
NN
What are the disadvantages of bilinear interpolation? - correct answer ✔✔- Gray values are
altered
- Some blurring (e.g. filter)
- Slower
Bicubic Convolution (CC) - correct answer ✔✔Assigns a value to a given cell in the output matrix
based on a 4x4 moving average calculated using a cubic polynomial
What are the advantages of bicubic convolution? - correct answer ✔✔- Very smooth image
- More visually pleasing
- Considers more of surrounding area than bilinear
What are the disadvantages of bicubic convolution? - correct answer ✔✔- Gray values are
altered
- Slower than bilinear interpolation
- Potential for values outside the range of the input data set
When should you not rectify to map coordinates? - correct answer ✔✔- When positional
information is not required (maybe only register image-to-image)
- Before (or after) spectral image classification (depends on resampling method)
What are the 3 types of radiometric corrections? - correct answer ✔✔1) Sensor's sensitivity
2) Sun angle and topography
3) Atmospheric correction (absorption, scattering)