4.1 Radiometric and geometric corrections

Question 1

Remote sensing data pre-processing

Answer 1

• Remote sensing data are often radiometrically degraded
•Radiometric corrections
–Cosmetic corrections: Line dropouts, Line striping, Random noises
–Sun elevation
–Haze removal
–Atmospheric correction

Question 2

Radiometric correction

Answer 2

Radiometric correction improve radiometrically degraded remote sensing
– Correcting for imperfections of the sensor
– Correcting for scene peculiarities and atmospheric disturbances
– Image enhancement

Question 3

Radiometric calibration and correction

Answer 3

• There are visible errors and noises in the raw data that need to be corrected

[DN]

• Sensor dependent
• Re scaling factors often found in the metadata

[At sensor radiance]

• Requires additional information such as Solar zenith angle, Earth-sun distance, solar irradiance
• Often provided in the metadata

[TOA radiance and reflectancce]

• Requires knowledge of atmospheric conditions and aerosol properties at the time of image acquisition

[Surface reflectance]

Question 4

Cosmetic corrections

Answer 4

• Correcting visible errors and noise in the image data.
– Missing lines: caused by failure of one of the detectors
– N Line striping: caused by the missing calibration of one of n detectors
– Random noise (spikes)

Question 5

Missing lines

Answer 5

• Missing lines caused by failure of one of the detectors
– Random or periodic
• Defective lines are detected by comparing the average of each line with the scene average
• Defective lines are replaced with average of proceeding or exceeding lines.

Question 6

N-Line striping

Answer 6

• The response of some detectors might drift with time. Therefore, every line may be brighter or darker than the other lines.
• Line striping is usually corrected using histogram matching or filters in Fourier domain.

• Histogram matching.
– Separate histograms corresponding to each detector.
– One detector response is taken as reference.
– Other responses are matched to the reference.

Question 7

Spike noise

Answer 7

• Individual pixels with values much higher or lower than surrounding pixels.
• Caused by temporary disturbance or errors during data transmission.
• Identify spikes: each pixel is compared with its neighboring pixels.
• Pixels deviating more than a specific threshold from their neighbors are spikes.
• Spikes are replaced by interpolated values.

Question 8

Sun elevation correction (1/2)

Answer 8

• Sun elevation changes depending on time of day and day of year.
• Solar irradiance varies with the seasonal changes in solar elevation angle and the distance between the Earth and sun.
• For comparison of images acquired at different time of day or different days of year, sun elevation correction is important.
• note:
• Irradiance that is received by the target is higher in summer compared to winter
• higher sun elevation gives better contrast

Question 9

Sun elevation correction (2/2)

Answer 9

• Sun angle normalization
DN' = DN/sin(a)
where
DN = raw pixel value 
DN' = Corrected pixel value 
a = sun elevation angle 

• Sun elevation is usually provided in the metadata
• Relative sun angle correction
– Stable land cover is used as reference
– Radiometric values of the image with lower sun elevation are adjusted to the other image

Question 10

Haze removal

Answer 10

• Haze correction removes sky radiance from the remote sensing data
– Sky radiance reduces image contrast.
– It has an additive effect, resulting in higher DN values (makes the image brighter but lower contrast).
– Sky radiance depends on wavelength.
– Haze removal is done for each band independently.

• Estimating the correction constant from the image
– When there is areas that should have zero reflectance (e.g. water in Infrared band)
– Smallest value in each band is taken as the correction constant

• From external information
– Additional information e.g. aerosol distribution and gas composition needed

Question 11

Atmospheric correction -Relative atmospheric correction

Answer 11

– Based on ground reflectance
– Approximate
– Assume an empirical linear relationship between TOA and ground radiance.
– Methods
• Two reflectance measurements
• Two reference surfaces

Question 12

Atmospheric correction -Relative atmospheric correction

Method: Two reflectance measurements

Answer 12

Procedure

– The image is corrected based on field radiance measurements.
– The output is an absolute atmospherically corrected image.

• Some bright and dark targets, each covering a few pixels in the image, are selected.
• During the satellite pass, the reflectance of targets is measured in the field, using a reflector-meter device.
• Surface reflectance is measured using portable radiometers at spectrum bands similar to those of the sensor.
• A linear equation between TOA radiance and surface reflectance is built for each band.
• The linear equation is applied to each band to obtain corrected image

Question 13

Atmospheric correction -Relative atmospheric correction

Method: Two reference surfaces (we want to compare the image from the same area but at a different time of atmospheric circumstances)

Answer 13

– The atmosphere of a reference image is imposed to another image.
– Relies on reflective invariant (the reflection of the target does not change through time) bright and dark areas in the image.
– Any difference in the reflectance of bright and dark areas are attributed to atmospheric conditions.
– The output is an image corrected relative to a reference image.
– Valid for comparing images.

Procedure
• At least one dark and one bright invariant areas in the image are selected (e.g. deep reservoir lakes, big asphalt areas, sandy beaches).
• The radiance difference between the areas in two images are calculated
• Based on the differences, a linear function is calculated.
• The linear function is applied to the second image.

Question 14

Atmospheric correction -Absolute atmospheric correction

Answer 14

• Radiative transfer models
– Can be used for computing either radiances (intensities) or irradiances (fluxes) for a variety of atmospheric and surface
– Use detailed scattering models of atmosphere from physically based models.
– They need a full description of atmospheric components at fixed altitudes e.g. MODTRAN ® and 6S
– A variety of input parameters are required:
§ Geometric conditions (sun elevation and sensor altitude)
§ Atmospheric model for gaseous components (Rayleigh scattering) e.g. H2O and O3
§ Aerosol models (Mie Scattering) e.g. type and concentration of aerosols like dust, salt, etc
§ Spectral bands and bandwidth