Tutorial on image registration

This tutorial will demonstrate how to register two images by aligning a map of Chernobyl to a Landsat image.

1. Open the files chernobyl1.tif.gz (a map of the Chernobyl area) and chernobyl2.tif.gz (a small version of a Landsat image of the same region from the U.S. Geological Survey).
2. Click on the edge of chernobyl2.tif to select it and move it so both images are visible.
3. Open the Image Registration Dialog and click on ``Img 1 Obtain landmks''. Then click on the Chernobyl map at the points labeled 1 to 5 (the two small lakes, the south end of the large island, end of the road, and the island). Press Space when finished.
4. Click on ``Img 1 Obtain landmks'' and click on the Landsat image at the points labeled 1 to 5. Press Space when finished.
5. Click on ``Edit Landmarks'' to view the data points. It should resemble the table below:

    # Ref. Ref.x Ref.y   Unk.  Unk.x  Unk.y  xmin1  ymin1  xmax1  ymax1
           96    86      -     206    93     0      0      0      0
           280   59      -     350    73     0      0      0      0
           275   295     -     349    266    0      0      0      0
           101   314     -     209    282    0      0      0      0
           223   185     -     305    172    0      0      0      0
   

6. Click on `Calculate warp vector map'. A blue-green image will be created.
7. Click `Warp image' to warp the Landsat image.
8. Click `Config..Configure' and set the Main Cursor to `multiple crosshairs' and verify that the registration points are at similar locations. If the crosshair cursor is too dark, click `Fcolor' and change the main drawing color.
9. After warping, the landmark points are automatically remapped. Click on `Calculate warp vector map' and `Warp image' again to perform a closer approximation. More repetitions can be performed until the two sets of registration points become identical. This usually occurs after two to three iterations.

$\parbox{1.6in}{ \raggedbottom \setlength{\lineskiplimit}{-99pt} \... ...feature are the same as the image coordinates of every feature on the map. }$