Faster and scalable parallel processing solution to remove visual obstacles from satellite imagery

The use of artificial satellites, especially? from ESA Copernicus program, created new opportunities for sciences geared toward studying phenomena that have a major impact on our planet, especially anthropogenic phenomena. In this way, accurate measurements and predictions could be made regarding the degree of pollution of land and water, the evolution of deforestation and desertification. However, in order to obtain relevant data from satellite imagery, they must be passed through a procedure to remove visual obstacles, such as clouds, shadows, and sometimes snow. An important drawback of filtering algorithms is the extremely low performance that makes some processing last from a few hours to a few days. This paper attempts to eliminate the major disadvantage of extensive data processing time by proposing a much faster and scalable parallel processing solution. The paper starts from the context setting and the theoretical description of the filtering algorithm used and the main optimization technique, then goes on to detail the actual implementation and ends with the exposition of the results obtained from the extensive qualitative validation and the measurement of the performance indices and efficiency.
Satellite imagery, Parallel processing, Scalability, Optimization, Landsat, Cloud detection algorithm