The first challenge in information extraction from optical satellite images of the Earth's surface is the collection of cloud free images, as one third of this data can be affected by cloud cover.
The first challenge in information extraction from optical satellite images of the Earth's surface is the collection of cloud free images, as one third of this data can be affected by cloud cover. Clouds dramatically affect the signal transmission in complex ways due to their different shapes, heights, and distributions and thus contaminate the data from land and water. Cloudy image restoration is a vital step in the remote sensing image processing chain. Correction of cloudy data can substantially increase the number of useable images and pixels available for later applications such as mapping land cover types and sea surface features. Cloud correction techniques strive to remove cloud effects, including cloud shadows, and recover the Earth's surface information in contaminated pixels. However, current methods often have limited success, especially when the imaged areas beneath clouds have complex ground characteristics. Most existing techniques are also narrowly focused on individual images rather than time series or multi-temporal data.
In this project, we aim to develop reliable methods for cloud effect removal. The project can focus on thin cluod data restoration, thick cloud data replacement, or shadow correction. These projects will involve cutting edge research with the development of the advanced digital image processing techniques, for example. data decomposition approaches, sparse representation techniques, and spectral unmixing.
School of Engineering & IT
Imaging | AI for Space