How can you make Image-to-Image Registration results exactly match the base image?

Note:  This Help Article may refer to an older version of ENVI and may not be relevant to more recent versions.

 

When you do a registration warp in ENVI, the default behavior is that ENVI will produce a registration result that includes the entire input image, regardless of where the input image overlaps with the base image. (The only exception to this rule is if you are using a triangulation warp model, in which case the coverage of the registered result is dependent on the location of the GCPs). In some cases you may want the coverage of the registration result to exactly match that of the base image.

For example, this can be very useful when you need to do Band Math using both the base and registered images, or if you want to do an Animation with registration results where all of the images will be resampled identically.

It is easy to set up the registration so that the output warped image will be clipped to match the coverage of the base image. In the last step before you begin the warp processing, you will be in the Registration Parameters dialog. Select the button in the middle of this dialog called Change Output Parameters.

Notice that the values for the upper left pixel (X0 and Y0) are in pixel coordinates relative to the base image -- that's why their values may appear unexpectedly large or small, or even negative! ENVI uses a coordinate system like this so that the registration result's pixel offsets (i.e., the Xstart and Ystart values in the ENVI header file) will allow you to use the Dynamic Overlay to compare the base and warp images. If you want the registration result to exactly match the coverage of the base image, change the upper left pixel to the value (1,1), and change the number of samples and lines to be the same size as the base image. For example, if the base image you're using has 356 samples and 464 lines, then you would change the Output Parameters to look like this:

However, remember that if your input image's coverage is smaller than that of the base image, you will end up with a result that has areas of 'no-data' where there is simply no coverage.