Filter sentinel 2 images that have empty areas

Question

When trying to download a clipped area from the sentinelhub API python library, it returns images from 2 different orbits: One that passes right on top (Right pic), and the image is full, and a orbit next to (Left pic), but its image is cut due to deformations.

Checking the metadata in the sentinelhub website, there is a parameter called no_data_pixel-percentage, but that parameter is not included in the metadata returned by the API. Does anyone have found a solution to filter the good ones?

Answer 1

sentinelhub Python package allows you to get a no-data (i.e. transparency) mask for each image. Example, how to do this, can be found in documentation, chapter 10.

Basically you have to specify custom_url_params={CustomUrlParam.TRANSPARENT: True} when you initialize your WMS or WCS request. Downloaded images will have one channel more then before and the last channel will have information about no data.

Once you have this channel you can decide which images to use. The package does not have a "maximal no-data percentage" parameter with which you could download only images that have low no-data percentage. However you can use time_difference parameter to merge images from multiple acquisitions into a single one as shown in documentation examples, chapter 8.