Home /Processing/Processes

Coregistration

An algorithm that improves the positional accuracy of imagery against a reference.

Overview

The process aligns a source image to a reference image. Its algorithm automatically detects and minimizes misalignments between features using rigid-body transformation. The results are GeoTIFF files of matching optical assets with improved positional accuracy.

Coregistration can be used as a preprocessing step before change detection, data fusion, and DEM generation.

Examples of input images:

A source image that hasn't been coregistered, with a highlighted reference area
A reference image with better positional accuracy, with a highlighted reference area

The results are the following:

  • A GeoTIFF of the coregistered source image
  • A GeoTIFF floating point displacement map of calculated misalignment after coregistration
  • A GeoTIFF heatmap for displacement, with warmer colors showing greater misalignments

A floating point displacement map and a heatmap won’t be generated in case of too few tie points in the input files.

A coregistered source image, with a highlighted reference area
A map of calculated misalignment overlaid on the coregistered source image
A heatmap for displacement overlaid on the coregistered source image

General

Specification Description
Provider Simularity
Process type

Enhancement from 50 credits per km2

Performance

The coregistration algorithm aligns images accurately, even if they have different resolutions or sensor types. It can handle clouds and land cover changes. It calculates sub-pixel shifts by analyzing local image patches, then aligns the images using rigid-body translation and rotation.

To get the best results, make sure that:

  • Both data items have matching geometric and radiometric processing levels.
  • The GSD of the two data items doesn’t differ by more than 25%.

Requirements for input imagery

Input data items must come from a CNAM-supported collection and be added to storage in 2023 or later.

CriteriaRequirement

Product type

Input data items must come from a multispectral collection.
Processing levelInput data items must be georectified or orthorectified.
Spectral bands

Input data items must have an asset with the same bands.
GeometryThe geometries of input data items must overlap by at least 20%.
Cloud coverageThe cloud coverage of input data items must be less than 25%.

Input parameters

Use the coregistration-simularity name ID for the processing API.

JSON
{
  "inputs": {
    "title": "Coregistering imagery over Berlin",
    "sourceItem": "https://api.up42.com/v2/assets/stac/collections/21c0b14e-3434-4675-98d1-f225507ded99/items/23e4567-e89b-12d3-a456-426614174000",
    "referenceItem": "https://api.up42.com/v2/assets/stac/collections/21c0b14e-3434-4675-98d1-f225507ded99/items/edeb6310-ea9a-4d8e-943d-11a5f3757824"
  }
}
Parameter Overview
inputs.title string | required

The title of the output data item.
inputs.sourceItem string | required

The absolute API path to the source data item.
inputs.referenceItem string | required

The absolute API path to the reference data item. The positional accuracy of the source image will be improved against this reference.

Learn more