Sensor calibration converts the sensor’s digital numbers (DNs) to at-sensor radiance above the atmosphere. A further radiometric adjustment accounts for the viewing angle and sun angle during acquisition to transform radiance values to top-of-atmosphere (TOA) reflectance. Therefore, the process requires sensor calibration information and telemetry data that satellite image providers deliver within the metadata.
DNs are raw sensor data without physical units. The sensor calibration information for converting the DNs to radiance are the calibration gain (cal_gain) and calibration offset (cal_offset) values. The sensor calibration uses linear function f(DN) = DN * cal_gain + cal_offset that multiplies the DNs of each pixel in each spectral band with their corresponding cal_gain and adds the corresponding cal_offset. The resulting at-sensor radiance image is the basis for the radiometric adjustment that uses information about the viewing angle and sun angle during acquisition to transform at-sensor radiance to TOA reflectance.
Sensor calibration obtains TOA reflectance and is a minimum requirement for performing band math calculations to derive spectral indices such as the normalized vegetation difference index (NDVI). Uncalibrated image data would arrive at NDVI values that are distorted because the cal_gain and cal_offset parameters for the involved spectral bands were not considered.