/

Super-resolution Sentinel-2

Runs a deep-learning based superresolution algorithm to create a Sentinel-2 image with 10 m resolution across all bands.


Introduction

For more information, please read the block description.

Block type: PROCESSING

This block takes an Sentinel-2 L1C/L2A MSI Full Scenes image and runs superresolution algorithm on it. This algorithm creates a 10m resolution for all the existing spectral bands within 20m and 60m resolutions by using a trained convolutional neural network.

This block works on a minimum image size of 192x192 pixels. This is equivalent to a minimum area of 4 km².

Supported parameters

  • bbox: The bounding box to use as an AOI. Will clip to scenes that intersect with this box. Use only bbox or intersects or contains.

  • intersects: A GeoJSON geometry to use as an AOI. Will clip to scenes that intersect with this geometry. Use only bbox or intersects or contains.

  • contains: A GeoJSON geometry to use as an AOI. Will clip to scenes that intersect with this geometry. Use only bbox or intersects or contains.

  • clip_to_aoi: When set to true, the area that defined in bbox, contains, or intersect for previous data block will be clipped for processing. Note that by default this parameter is false which means that the whole scene will be processed.

  • copy_original_bands: If set to true, the original 10m resolution band will also be included in the output image.

Example parameters using the Sentinel:2 L1C MSI Full Scenes block as data source and then applying s2-superresolution clipped to the specific AOI:

{
  "sobloo:s2:l1c:fullscene:1": {
    "ids": null,
    "time": null,
    "limit": 1,
    "intersects": {
      "type": "Polygon",
      "coordinates": [
        [
          [
            55.341868,
            25.240834
          ],
          [
            55.350924,
            25.240542
          ],
          [
            55.350666,
            25.233244
          ],
          [
            55.342362,
            25.233322
          ],
          [
            55.341868,
            25.240834
          ]
        ]
      ]
    }
  },
  "s2-superresolution:1": {
    "intersects": {
      "type": "Polygon",
      "coordinates": [
        [
          [
            55.341868,
            25.240834
          ],
          [
            55.350924,
            25.240542
          ],
          [
            55.350666,
            25.233244
          ],
          [
            55.342362,
            25.233322
          ],
          [
            55.341868,
            25.240834
          ]
        ]
      ]
    },
    "clip_to_aoi": true,
    "copy_original_bands": false
  }
}

Output format

GeoTIFF format. Also in the output file all the spectral bands for 20m and 60m will be in 10m spatial resolution.

Capabilities

Input

raster
up42_standard
bands
[
  "coastal",
  "blue",
  "green",
  "red",
  "rededge",
  "rededge2",
  "rededge3",
  "nir",
  "nir2",
  "watervapour",
  "swir",
  "swir2",
  "swir3"
]
dtypeuint16
formatSAFE
sensorSentinel2
resolution10
processing_level
{
  "or": [
    "l1",
    "l2"
  ]
}

Output

To know more please check the block capabilities specifications.

Download example output

You can create example output to use when testing processing blocks built to work with this data by running the block in a workflow via the console, and downloading the results in the job overview.