Written and maintained by CASRAI Editorial Board
Last updated
To georeference an image in QGIS, open Layer > Georeferencer, load the scanned map or photograph, click ground control points (GCPs) that you can match to known coordinates, choose a transformation type, and press Start georeferencing. The result is a raster that sits at the right place on the map. The choice that matters most is the transformation, because each type needs a different minimum number of GCPs and only two of them can produce a world file instead of a new image.
What georeferencing in QGIS is used for in research
Georeferencing gives an image that has no location, such as a scanned historical map, an aerial photograph or a field plan, real-world coordinates so it can be overlaid on other spatial layers. The QGIS tool for this is called the Georeferencer. Researchers use it to place archival maps next to modern boundaries, to align photographs with survey data, and to prepare rasters for analysis. The same dialog also accepts vector layers, which is useful when a digitised layer is offset or stored in an unknown reference system.
Because the position of every pixel depends on the points you choose, georeferencing is a measurement step. Record which GCPs, which transformation and which resampling method you used, in the same way you would document any other processing step (see computational reproducibility). Georeferencing is also the step that produces the location information that later becomes metadata for the dataset.
Install or access QGIS (official source only)
Download QGIS only from the project’s own download page, qgis.org/download. Checked on 2026-10-29, the page lists desktop builds for Windows, macOS, Linux and BSD, plus containers, and offers an OSGeo4W online installer for Windows. It offers a Long Term Release (LTR 3.44) and a Latest Release (4.2) for download. Older tutorials that send you to a Plugins or Raster menu describe the earlier Georeferencer GDAL plugin; in the current documentation the tool is under Layer > Georeferencer. For a broader tour of the program, see how to use QGIS software.
Step by step: how to georeference an image in QGIS
The steps below follow the QGIS documentation page “Georeferencer” (checked 2026-10-29: docs.qgis.org). Button names are quoted as they appear there.
- Choose Layer > Georeferencer. Settings are under Settings > Configure Georeferencer (Ctrl+P).
- Click Open raster and select the image. It appears in the dialog’s work area.
- Click Add GCP Point, then click a feature on the image that you can identify on a georeferenced source, such as a road junction or a building corner.
- Give that point its real-world coordinates. You can type X and Y and choose the point’s CRS, or use the pencil From map canvas button to pick the position from a georeferenced layer already loaded in the main QGIS window; the CRS is then set automatically.
- Repeat for more points. Use Move GCP Point to correct a misplaced point and Delete GCP Point to remove one. Link QGIS to Georeferencer and Link Georeferencer to QGIS keep the two map extents in step while you work.
- Click Save GCP Points As so you can reload the points later with Load GCP Points.
- Open Transformation settings, choose the transformation type, resampling method and Target CRS, and set the output file.
- Click Start georeferencing. If you tick Load in project when done, the output is added to the map canvas automatically.
Georeferencer in QGIS for vector layers
Click Open vector instead of Open raster. Turn on snapping with the Snapping Type button if you want points to snap to features, add GCPs in the same way, then set the transformation type, Target CRS and output path. The default output name gets a _modified suffix.
How many GCPs do I need?
The minimum depends on the transformation type. The QGIS documentation also advises using at least four points and notes that more coordinates generally give better results.
| Transformation | Minimum GCPs |
|---|---|
| Linear | 2 |
| Helmert | 2 |
| Polynomial 1 | 3 |
| Polynomial 2 | 6 |
| Polynomial 3 | 10 |
| Projective | 4 |
| Thin Plate Spline | 10 (usually more are needed) |
With more GCPs than the minimum, the fitted result matches the points only approximately. Thin Plate Spline matches them exactly, but the documentation warns that it can distort the area between nearby points if there are registration errors.
Which resampling method?
The documentation lists five methods: Nearest neighbour, Bilinear (2×2 kernel), Cubic (4×4 kernel), Cubic B-Spline (4×4 kernel) and Lanczos (6×6 kernel). Record the one you choose alongside the transformation.
Output and how to read it
- Georeferenced raster. Unless a world file is created, an output raster is required; its default name is the input name plus
_modified. Raster creation options control how the file is structured and compressed. - World file. Create world file is available only with the Linear or Helmert transformation, because those do not transform the pixels. The Output raster field is then disabled.
- GCP file. Points are saved as
[filename].points, usually beside the image, with four values per row: mapX, mapY, pixelX, pixelY. Keep this file with your project data so the georeferencing can be repeated or audited. - PDF map and PDF report. The report contains the transformation parameters, a residuals image and the GCP list with RMS errors. Under Configure Georeferencer, the Residual Units setting shows residuals in pixels or in map units.
- Other options. Set Target Resolution defaults to 1 for horizontal and vertical resolution, and Use 0 for transparency when needed makes 0-value pixels transparent.
The report lists the RMS error for each GCP, so you can see which points fit least well and adjust them with Move GCP Point or Delete GCP Point before running again.
Troubleshooting
The transformation will not run or looks wrong
Check the count against the table above. A Polynomial 2 transformation needs 6 GCPs and Polynomial 3 needs 10; with fewer points, choose a lower-order type or add points.
The image is stretched between nearby points
This is the behaviour the documentation describes for Thin Plate Spline when GCPs contain registration errors. Review the points, or use a polynomial transformation with more GCPs than its minimum.
No output raster field, or it is greyed out
With Create world file ticked, the Output raster field is disabled. Untick it if you want a new image, or use Linear or Helmert if a world file is what you need.
I cannot pick coordinates from the map
From map canvas reads positions from a georeferenced layer in the main QGIS window, so load one first. If the Georeferencer window is in the way, tick Automatically hide georeferencer window; the documentation notes this only applies when the window is not docked.
The result does not appear in QGIS
Tick Load in project when done before you start, or add the output file as a normal raster layer afterwards.
Free alternatives for georeferencing images
| Tool | How it works | Source |
|---|---|---|
| QGIS Georeferencer | Desktop dialog with GCP table, transformation settings and PDF report (this guide) | QGIS documentation, checked 2026-10-29 |
GDAL gdalwarp |
Command-line tool; -order takes 1, 2 or 3 for the polynomial order, -tps selects Thin Plate Spline and -t_srs sets the target spatial reference |
GDAL documentation, checked 2026-10-29 |
| Map Warper | Online georeferencing (“georectification”) service for images and maps; its about page describes it as an open source application and names Tim Waters as the contact | Map Warper about page, checked 2026-10-29 |
ArcGIS also georeferences images; for what that program covers, see what ArcGIS is used for.
FAQ
How do I georeference an image in QGIS?
Open Layer > Georeferencer, load the image with Open raster, add at least the minimum number of GCPs for your transformation, set the transformation and output in Transformation settings, and click Start georeferencing.
Where is the Georeferencer in QGIS?
Under Layer > Georeferencer in the current documentation. Its settings are at Settings > Configure Georeferencer.
Does georeferencing in QGIS work for vector data?
Yes. Use Open vector and follow the same GCP workflow; the output is written with a _modified suffix by default.
How many control points does QGIS need?
Two for Linear and Helmert, three for Polynomial 1, four for Projective, six for Polynomial 2, ten for Polynomial 3 and at least ten for Thin Plate Spline. The documentation advises at least four points in general.
Can I reuse my control points?
Yes. Save them with Save GCP Points As and reload them with Load GCP Points. They are stored as a .points file with mapX, mapY, pixelX and pixelY columns.
Sources (official documentation)
- QGIS Documentation, “Georeferencer” (current “latest” manual): docs.qgis.org, checked 2026-10-29.
- QGIS download page: qgis.org/download, checked 2026-10-29.
- GDAL documentation, gdalwarp: gdal.org, checked 2026-10-29.
- Map Warper, about page: mapwarper.net/about, checked 2026-10-29.
