Registration of Digital Images by
Warping Control Points and Curves

To view the demonstration, click here

An algorithm has recently been developed by Professor M.J.D. Powell of Cambridge University for registering (overlaying) two images of the same scene by transforming control points and curves in the first image into target control points and curves in the second image. This so-called Spider Warping algorithm employs thin plate spline functions to map the chosen control points exactly onto their target values, while mapping digitized versions of the chosen curves to within a user-defined deviation from their target curves. The resulting mapping represents the smoothest possible transformation which exactly overlays the control points, while approximately overlaying the chosen curves. The Spider Warping algorithm has several applications, including medical diagnosis, synthesis of cartographic data, and detection of mines in shipping channels. Its ability to deal effectively with "spider-like" images such as river networks, underwater ridge lines and other curvilinear features, distinguishes this high-accuracy warping algorithm from other registration methods.

Potential users of this Spider Warping algorithm should contact BCS for further information about currently available software and prices.

Spider Warping Demonstration

Note: The following demonstration requires a display resolution of 800 X 600 pixels or more.

Our demonstration of the Spider Warping algorithm consists of four images which can be simultaneously shown in overlapping or flickering mode. On the top-left, the original images are shown: the first image (the one to be warped) is in green and the second image is in white. On the top-right, the first image has been warped so that its control points are mapped exactly onto the corresponding (red) control points in the second image. (In flickering mode, place your mouse pointer on any of the 14 red points to verify visually that this warp exactly overlays the control points.)

On the bottom-left, the Spider Warping algorithm has been applied with three yellow "spider legs" and the same 14 control points. (In flickering mode, observe how well these three corresponding pairs of legs now match, while other pairs of legs are still not close matches.) Finally, the bottom-right images illustrate the stabilizing effect of including more yellow "spider legs"

To view the demonstration, click here.