Register Command

A new command was added to the Stacks menu that will translate and rotate the slices in a stack into alignment based on fiducial points. This command was written by Michael Castle  (mike.castle@med.umich.edu) of the University of Michigan Mental Health Research Institute  (MHRI).

The Register command offers two interfaces.  The more intuitive of the two requires the user to click on visible landmarks in a reference slice, then click on corresponding fiducial points in the other slices in the stack.  The code then lines up the slices based on these points.  The user may easily throw out unreliable points in non-reference slices and correct input errors.  This registration interface may only be used in cases where all fiducial points are visible on screen in the captured slices.

The second interface is much more difficult to use, but it works regardless of whether or not fiducial marks are included within the borders of the captured slices.  Here, the user provides fiducial data in a text file (which is much more convenient now that NIH Image can open and edit text files!).  This data consists of relative coordinates of fiducials from each slice as read from a microscope digitizer (or similar device) as well as the necessary information to map the coordinate system of the microscope digitizer to the coordinate system of windows in NIH Image.

As mentioned before, registration may be performed either by clicking on a set of fiducial points in each slice or by reading such data from a text file. The on-screen clicking method is self-explanatory, but the text file method is somewhat less clear and need only be used when the fiducial points do not lie within the boundaries of the captured slices. 

The text file method uses relative coordinates of fiducial marks as recorded with some device such as a microscope digitizer. Let us call these coordinates "stage" coordinates. It also uses screen coordinates of fiducials in Image. Before registration may take place, stage coordinates must be mapped to the NIH Image screen coordinate system. To do this, NIH Image reads the following, tab-delimited information from a text file. In every case, x-coordinates are followed by y-coordinates.

line 1: the screen coordinates of a point which is fixed for every image captured. In the sample fiducial file, this is the point which corresponds to the crosshairs on the microscope at capture time. This is the "image center" point.

line 2: the location, in screen coordinates, of two fixed points in an image. 

line 3: the location, in stage coordinates, of these same two fixed points. 

line 4: (and each subsequent line) the locations, in stage coordinates, of the "image center" and at least two fiducial points for an image to be registered. 

Obviously, the stage coordinate system may not change in rotation, translation, or scale in the middle of a set of images being captured, so it is advisable to record stage coordinates at the time of image capture (without changing camera settings or the position of the object whose image is being captured). 

Lines 4-n must entered in the same order as the slices in the stack. It doesn't matter which one is the reference image. 

Where fiducial coordinates are unavailable or unreliable, use the number 10000 for both the x and y stage coordinate in the file. 

Each slice must be represented by the same number of fiducial coordinates in the data file.

As you can see, using the text file method introduces many possible sources of errors, so results are not usually as good with this method as with the on-screen method.

Currently, no more than 12 fiducial points may be specified for each slice.