GIS and data analysis
Since they can process, cloud-store, and make available a large amount of data, Geographic Information Systems (GIS) enable complex spatial analysis and prove to be extremely powerful instruments for “decision making” analysis in a lot of different scenarios.
The need for Georeferencing
To perform analysis into a GIS environment, however, it is necessary to match all the input data into the same geographic reference system. Georeferencing is the process of taking a digital image and adding geographic information to it so that GIS or mapping software can “place” the objects represented on it in their real-world location.
In other words, the objective of georeferencing is to assign to a set of pixels in the digital image (Ground Control Points) their geographic coordinates in a given coordinate system.
Geometries and accuracy
Objects in images can be assigned a geo-referred geometries. However, the georeferencing of an image dataset is not always precise, and depending on the application in change detection, such as measurement of seismic ground deformation, ice flow, or dune migration, the required accuracy may vary, but generally, it should be less than 1 meter .
The difference in source and time may cause displacements
In GIS environment, images (matrix) come from different kinds of sources: sensors on board of airplanes, drones, or satellite platforms. Moreover, images can refer to different dates: this is a widely used procedure for change detection monitoring on the terrestrial surface.
The differences in the source, time, and viewpoint in the image acquisition can introduce some sub-pixel geometry displacements between images. Displacements can occur even between images captured from the same sensor.
Co-registration and why it is important
Therefore, it is crucial to make images provided by remote sources useful for mapping . When working with two or more images datasets together, indeed, they all must match the same set of geometries in the same geographic reference system.
Co-registration in image processing is a procedure that helps minimize shifts between images at a sub-pixel scale. The reason why it is fundamental for remote sensing applications such as environmental mapping, change detection, mosaicking, or image fusion  is that by minimizing the displacement between image datasets, the co-registration process avoids errors at further analysis.
The co-registration process
The process of co-registration, generally speaking, fixes the deformation or distortion of an image with respect to a reference image. The reference image is an ortho-rectified and geometry-corrected image. It means that it was georeferenced from control points collected on the ground and also geometry corrected based on the terrestrial topography information (Digital Elevation Model – DEM).
For being co-registrated, the images must be projected and resampled onto the same reference system. The general process of the co-registration procedure takes both “projected” and “resampled” images (reference and target image), calculates the measurement of displacement in X and Y of both images, and corrects those displacements on the target image.
The displacement between images gets calculated by using image registration techniques, which are mainly of two types: intensity-based and feature-based . The first identifies the similarity of the pixel gray-values that appears on the reference and target image. The second technique detects the position of a ground object to distinct the feature in both reference and target images. After identified those shifts between images, the correction gets done by warping the target image, a mathematical transformation based on the displacement values.
From what we said so far, it emerges that GIS and its related instruments can generate very accurate and reliable datasets from large and diverse kinds of sources. Georeferencing and co-registration techniques are tools fundamental for reaching a high accuracy on the GIS spatial analysis.
 ‘Geographical information systems and science’ by Wiley, New York (2005)
 ‘Automatic and precise orthorectification, coregistration and subpixel correlation of satellite images, application to ground deformation measurements’ by California Institute of Technology (2007)
 ‘Georeferencing multi-temporal and multi-scale imagery in photogrammetry’ by University of Technology – Espoo, Finland(2008)
 ‘The impact of misregistration on change detection’ by University of Maryland, USA (1992)