BearFruit1 image registration

To correct for differences in image registration from wavelength-to-wavelength, we used a global affine motion estimation algorithm to compute transformations that register each image with the 550 nm image. We used code provided by Heeger that implements an algorithm described by Bergen et al. [8]. (For more information on the registration software, go here.) We estimated the transformations from a 512 by 512 pixel subimage that contained the resolution target in the upper center of the images, then applied these transformations to the whole images. We did this because the image registration algorithm, which was not designed to handle color, works robustly across wavelength only if applied to image regions that contain predominantly luminance modulation.

The two sets of images below illustrate the image registration quality for the BearFruit1B image. The top left image shows a close-up of the resolution target from the BearFruit1B image without any registration. If you look closely, you can see a great deal of red/green chromatic fringing along the black lines. This fringing is largely gone in the bottom left image, which shows the same region after registration

The registration transformation was computed using the resolution target and is not as effective in other image regions. The smaller images on the right below show the comparison for the grid target at the bottom of the BearFruit1B image. Here registration reduces but does not eliminate the fringing.

There are a number of approaches to improving the registration quality. Get in touch if you'd like to help.

The brightness and contrast of image regions below were adjusted to try to make the effects more clearly visible. The same transformations were applied to each image region. The speckle pattern you see in the background of the images is due to the dithering algorithm used to convert these images to GIF format and is not present in the originals.

Before registration

After registration

Author: David Brainard,
August 29, 1997 David Brainard