Forest Cover Warping

To view the demonstration, click here.

BCS has developed fast and effective algorithms for warping digital images, such as those representing forest cover. These techniques are based on interpolating a number of corresponding control points in the two images using minimum-curvature thin plate spline (TPS) functions. The resulting mapping represents the smoothest possible transformation which exactly overlays the control points. BCS uses special-purpose algorithms to stabilize the computation of the TPS coefficients and to allow the TPS functions to be evaluated very rapidly over a grid, allowing the practical application of TPSs to solve forest cover and other warping problems. We have also developed techniques for automated selection of control points, thereby greatly speeding up the definition of the control points required for the warping functions.

In the forest cover warping demo, images of four different locations are shown. For each location, two images were obtained: the first (source) from an aerial photograph, and the second (target) from an orthomap. Each source image was then warped onto the target image using specified control points. For each location, the demo shows, using flickering, the source/target image pair before warping (in the (a) series) and the source/target image pair after warping the source to the target (in the (b) series).

Note: These are 640x480 GIFs and 2-frame animated GIFs, and may take several minutes to load with slow internet connections. Once the images are loaded into your cache, the animations should run smoothly.

To view the demonstration, click here.