Image pre-processing focuses on transforming the electrical signal measured by a sensor to a processing level at which pixel values can be used for the next information extraction step. Therefore, pre-processing operations involve the removal of errors encountered while collecting remotely sensed data to get as close as possible to the true radiant energy and spatial characteristics of the study area at the time of data collection. Different sensor type (optical, radar, lidar) require different processing levels
The most common image pre-processing procedures include:
(1) Radiometric calibration involves the transformation of Digital Numbers (DN) to physical unit: radiance/reflectance. Radiometric calibration can be done before the launch of a satellite sensor, i.e. pre-launch calibration, or after launch. In the second case, the calibration is performed on-board or by comparing ground measurements with satellite radiance. Through radiometric calibration various scene illumination procedures such as sun elevation correction or earth-sun distance correction are applied. Furthermore, image noises caused by striping or line drop as happened in case of Landsat TM7 due to failure of the Scan Line Corrector (SLC) are also corrected using specialized procedures.
(2) Atmospheric correction accounts for two main processes: scattering and absorption. Scattering represents a disturbance of the electromagnetic waves caused by rayleight scattering (caused by very small particles such as the air molecules), mie scattering (caused by aerosol particles) and non-selective scattering (dust, smoke, rain etc.). Absorption occurs when the electromagnetic energy is absorbed by the atmospheric components. Therefore, atmospheric windows have to be removed before using the satellite images in the next processing steps. Atmospheric corrections can be carried out either using simple statistical methods or complex radiative transfer based methods
(3) Geometric correction is required to remove the distortions caused by the Earth curvature, Earth rotation, panoramic distortion due to the field of view of the sensor and the topography of the terrain. Geometrics distortions are corrected using Ground Control Points (GCP) and a Digital Elevation Model (DEM). In case of airborne images, additional distortions caused by variations in the platform altitude or velocity might occur.