Image Correction in Remote Sensing
Understanding interference and distortion in remotely sensed data
For optical imagery, the most scientifically useful quantity is 𝜌, the percentage of incident radiation (𝐼) that is reflected from the surface and detected by the sensor (radiant flux 𝑅). Conceptually, this is the fraction of the Sun’s light that is reflected by the Earth’s surface.
𝜌′=𝑅/𝐼
For a given spectral range, 𝐼 can be estimated directly from a reference Solar irradiance spectra. 𝑅 is generally calculated directly from the raw sensor data (e.g. digital number), using the sensor specific conversion formula. To correct for the illumination angle of the sun (𝜃), 𝜌′ can be adjusted based on a reflectance model. The simplest model, Lambertian, removes illumination effects using a cosine correction:
𝜌=𝜌′/𝑐𝑜𝑠(𝜃)
The imagery is now considered corrected to ‘top of atmosphere’ (TOA), meaning that the data are adjusted to model the reflectance measurement at the top of the atmosphere. At this point the imagery needs to be corrected for the effects of terrain (orthorectification) and distortion resulting from the signal passing through the atmosphere (atmospheric correction). Often the illumination correction is done at the same time as the terrain correction, as the illumination angle of each pixel may be different in rugged areas.
All imagery offered through the Descartes Labs data refinery has undergone some level of data correction described in this section, though the processing levels vary by sensor and band. All products have been georeferenced, optical imagery has been converted from raw digital numbers to top of atmosphere reflectance.
Orthorectification
Orthorectification is the process of removing topographic, illumination, and perspective effects from an image. When an image is collected, terrain (e.g. mountains and valleys) can cause some areas to receive different amounts of light, or to appear offset from their actual position. Orthorectification corrects these irregularities, using a digital elevation model of the area to normalize illumination across the scene and to remove distortions due to the terrain.
Atmospheric Correction
The atmosphere has an inherent influence on almost all remote sensing data. Atmospheric conditions can change the amount of electromagnetic radiation that reaches the Earth’s surface, and distort or diminish that radiation as it travels back through the atmosphere to a sensor. Atmospheric particles, the angle of the sun at the time the sensor was triggered, and the weather on a given day will all greatly impact the quality of an image. When a project requires utilizing imagery acquired under different times and conditions, it is essential to remove the effects of the atmosphere. When the effects of the atmosphere have been removed, the imagery has been modeled to surface reflectance, or the reflectance at which the sensor would record if it were just above the surface of the Earth.