Skip to main content
Fig. 3 | Journal of Big Data

Fig. 3

From: Scalable approach for high-resolution land cover: a case study in the Mediterranean Basin

Fig. 3

a, d, g True Colour images (TCI) of similar dates, b, e, h original cloud masks using Sentinel-2 SCL cloud-related pixels, c, f, i post-processed cloud masks. j) is the composite of the three images using the original S-2 cloud masks, and k is the composite using the dilated cloud masks. Both composites show in red the no-data pixels, where data is not available in any of all original images. Marked with circles are some examples where the post-processed masks improve the composite. Note that j is problematic to use to obtain validated data, as pixels near no-data are probably corrupted. This problem is solved in k as corrupted pixels are not used for the arithmetic median. Also, the resulting composite k is much more precise than j, more noticeable when there are few products (2–3) available for making the composite

Back to article page