Coregix

Pairwise raster coregistration for geospatial imagery.

Coregix is a Python package and command-line tool for coregistering one geospatial raster to another. It estimates a geometric transform from a source raster to a reference raster, applies that transform to the source image, and writes a coregistered GeoTIFF while preserving geospatial metadata and multi-band outputs.

Coregix is intended for residual image-to-image alignment: cases where rasters are already georeferenced and close to alignment, but still visibly offset at the pixel level. It does not replace orthorectification, sensor model correction, or CRS reprojection.

Registration Model

The default registration is area-based. Coregix estimates a translation followed by a rigid transform using mutual-information optimization, then applies the result to all source bands.

Outputs can be written on either the source-raster grid or the reference-raster grid. Keeping the source grid is useful when the source image belongs to an existing stack; writing to the reference grid is useful when preparing matched rasters for comparison.

Input Expectations

Coregix expects:

• source and reference rasters in the same CRS
• enough overlapping valid pixels for registration
• useful raster masks or nodata values for invalid regions
• imagery that is close enough for translation/rigid registration to converge

Large rasters can be processed in chunks. Edge cleanup is available for outputs with invalid border artifacts after resampling.

First Alignment

Use --moving-image for the source raster and --fixed-image for the reference raster:

vhr-align-image-pair \
  --moving-image /path/to/source.tif \
  --fixed-image /path/to/reference.tif \
  --output-image /path/to/aligned.tif

The same alignment can be run from Python:

from coregix import align_image_pair

result = align_image_pair(
    moving_image_path="/path/to/source.tif",
    fixed_image_path="/path/to/reference.tif",
    output_image_path="/path/to/aligned.tif",
)

print(result.output_image_path)

Next Steps

• Installation: set up the package and documentation environment.
• Concepts: read about grids, masks, edge proxies, and chunking.
• Align Image Pairs: run the CLI or Python API.
• Large Rasters: use chunked transform application.
• Edge Trimming: clean invalid border artifacts.
• API Reference: inspect generated API documentation.