Image Registration Tool

The Image Registration tool is used for transforming the layer to make it match a given reference image. For doing so, the tool calculates the difference of gray image equivalent of the current layer to that of the reference image and tries to reduce their difference by moving, rotating and scaling the layer. This process is repeated many times until a desired accuracy has been achieved.

Following parameters are used in image registration:

Image Registration Dialog
Reference Image This is the image against which the layer will be transformed.
Transformation Model This parameter specifies how to transform the image. Shift and Rotate instructs the tool to only use a rigid body transformation, which is a combination of translation (shift) and rotation (preserving all lengths and angles, 3 degrees of freedom). Affine Transformation is a combination of rotation, scaling or shear, and translation (6 degrees of freedom).
Matching Area This parameter specifies whether to take into account the whole layer or only the selected parts of it as match criteria for calculating the transformation parameters.
Accuracy This is the threshold indicating the desired accuracy of the transformation parameters. It is given as an integer indicating the number of digits behind the decimal point (e.g., 6 indicates an accuracy of 10-6). The tool will stop iterating as soon as the (normalized) difference between the transformation parameters in two subsequent steps becomes smaller than this value.
Iterations This is the number of steps after which the tool will stop unconditionally. It is used for terminating the registration process in non-convergent situations. The layer remains unchanged if the process stops in this way, i.e., before reaching the specified accuracy. Another means for stopping the process is to press the Cancel button, which will also leave the layer unchanged.
Interpolation This parameter specifies the interpolation method used for rendering the transformed image:
None No interpolation (take the value of the nearest neighbor)
Linear Bi-linear interpolation
Cubic Bi-cubic interpolation
Sinc (Lanczos) Lanczos interpolation
Clipping This parameter determines how to clip the transformed image:
Adjust Enlarge the layer to contain the transformed image
Clip Delete all parts of the transformed image which exceed the image boundaries
Crop Crop the layer to contain only the non-empty parts of the transformed image
Crop with Aspect Same as Crop, but conserving the aspect ratio of the original image