Image registration is a process of aligning two images acquired by the same/different sensors, at different times or from different viewpoint. Many image processing applications like remote sensing for change detection, estimation of wind speed and direction for weather forecasting, fusion of medical images need image registration.
To register images, we need to determine geometric transformation that aligns images with respect to the reference image. The most common transformations used are:
rigid (RST),
affine,
perspective.
The parameters of the geometric transformation are calculated based on a given set of tie-points.
The tie-points can be obtained manually or automatically as follows:
Manual method: in this case the algorithm will generate the registered image based on manually selected tie-points,
Automatic method: in this case the tie-points are automatically generated by other algorithm comparing the input image with another reference image. The generated tie-points are later used by the register algorithm to generate the registered image.
It is accessible
through:
Raster Processing >
Register...
This wizard consists of the following steps:
On the List of Layers select the raster layer to use as a reference. Optionally use Filter By Name field giving part of the layer name to help find the layer in the list.
Press Next to go to next step or Cancel to close the dialog.
On the List of Layers select the raster layer to be adjusted. Optionally use Filter By Name field giving part of the layer name to help find the layer in the list.
Press Next to go to next step or Cancel to close the dialog.
The tie-points interface is composed of three components (Tie Points, Information and Reference/Adjust images) and several related tools. The details of each component are described bellow:
Tie Points - this component is used to handle the tie points and shows a group of tools used to add, update, select, remove, and also for automatic acquisition. The tie points are also presented are the component Reference/Adjust images area.
Tie-points can be ordered by ID, error (maximum of direct and inverse mapping error), type (manual or automatic) and coordinates (X1,Y1 are reference image columns/rows and X2,Y2 are adjust image columns/rows). To do that just click on a tie-points table column title.
To handle the tie points use the tools described bellow:
This tool is used to update the tie points values if some of the geometric parameters at the Informations component were changed.
This tool is used to add manually a tie point through the keyboard, giving the X and Y coordinate for both images reference and adjust.
This tool is used to select all tie points to execute the registration.
This tool is used to unselect all tie points.
This tool is used to remove all selected tie points from the list.
This tool is used to acquire automatically tie points over the the images used as reference and adjust.
This tool is used to finish the tie point acquisition step and shows the next step of the registration method.
Informations - (General and Options tabs) - it is used to define geometric parameters and also all options of specific parameters.
2.1 In General Tab the following information are required:
Reference Band: Band of reference image to be used in register operation.
Adjust Band: Band of adjust image to be used in register operation.
Input: Reference image metadata.
SRID: Information about spatial reference system from image.
Resolution X: horizontal resolution of the image pixels.
Resolution Y: vertical resolution of the image pixels.
Output: Adjust image metadata.
SRID: Information about spatial reference system from image.
Resolution X: horizontal resolution of the image pixels.
Resolution Y: vertical resolution of the image pixels.
Tie points number: Number of acquired tie points.
Transformation Error: Transformation error (The maximum of direct and inverse mapping errors).
All: For all tie points acquired.
Selected: For selected tie points.
Unselected: For unselected tie points.
2.2 In Options Tab it is possible to define
specific parameters for each type of geometric
transformation available. See the General, Moravec and Surf
tabs:
'
2.2.1 Options > General
2.2.2 Options > Moravec
Correlation Window Width: used to correlate points between the images (minimum 3, default: 11).
Gaussian Filter Iterations: the number of noise Gaussian iterations, when applicable (used to remove image noise, zero will disable the Gaussian Filter, default:1).
Minimum Correlation: the minimum acceptable absolute correlation value when matching features (when applicable), default:0.5, valid range: [0,1].
Moravec Window Width: the Moravec window width is used to locate candidate tie-points (minimum 3, default: 5 ).
2.2.3
Options > Surf
Maximum Euclidean Distance: the maximum acceptable euclidean distance when matching features (when applicable), default:0.5, valid range: [0,1].
Octaves Number: the number of octaves to generate, when applicable (default: 3, minimum:1).
Scales Number: the number of sub-sampling scales to generate, when applicable (default:4, minimum:3).
2.3 In Info Tab current transformation details is displayed:
Transformation Name: Current transformation name.
Total tie-points number: Total tie-points number.
Selected tie-points number: Current selected tie-points number.
Unselected tie-points number: Current unselected tie-points number.
Current mapping errors: For all tie-points. Measured in pixel units. It is important to notice that the reference and adjust images can have distinct pixel sizes.
Maximum Inverse Mapping error (all tie-points).
Maximum Direct Mapping error (all tie-points).
Maximum Inverse Mapping error (selected tie-points).
Maximum Direct Mapping error (selected tie-points).
Affine transformation details: For all tie-points. Measured in pixel units – It is important to notice that the reference and adjust images can have distinct pixel sizes.
Affine translation X (Direct - all tie-points).
Affine translation Y (Direct - all tie-points).
Affine scaling factor X (Direct - all tie-points).
Affine scaling factor Y (Direct - all tie-points).
Affine skew (Direct - all tie-points).
Affine squeeze (Direct - all tie-points).
Affine scaling (Direct - all tie-points).
Affine rotation (Direct - all tie-points): Measured in radians.
Affine translation X (Inverse - all tie-points).
Affine translation Y (Inverse - all tie-points).
Affine scaling factor X (Inverse - all tie-points).
Affine scaling factor Y (Inverse - all tie-points).
Affine skew (Inverse - all tie-points).
Affine squeeze (Inverse - all tie-points).
Affine scaling (Inverse - all tie-points).
Affine
rotation (Inverse - all tie-points): Measured
in radians.
Reference/Adjust images area: this component is used to show the tie points acquired in both images. There is one group of tools available to handle image composition. Using these tools it is possible to see the images in monochromatic, color composition, in red, green or blue channel. It is also possible to change the original color composition.
Also there is a group of tools
(vertical) to handle zoom in
, zoom
out
,
pan
,
recompose
,
manual collection of a tie point
over the images. To help collect precise tie points there is a
tool to display
an
small area beside both images highlighting and zooming the region
around where the cursor is pointing to. For the details on how to use
these tools see here.
Raster Info - First press and inform the folder where the resulting file will be saved.
Name - inform the raster name.
Extra Parameters - if there are some, see the details on how to inform then here.
Press Finish to save the resulting registered raster or Back to go to the previous wizard page.
Hint: Add the resulting raster to the TerraView project and draw it together with the raster used as reference.