Scale-based description and recognition of planar curves

A method of finding points of inflection on a planar curve at varying levels of detail and combining them to obtain a representation of the curve invariant under rotation, uniform scaling and translation and an algorithm to match two such representations are developed. This technique is applied to register a Landsat aerial image of an area (corrected for skew) to a map containing the shorelines of the same area. The shorelines are extracted from the Landsat image by forming a histogram of the gray-level distribution of pixels in the image and finding the land and water peaks in that histogram. The value at the trough between the two peaks is then used to threshold the image. Contours of dark regions in the resulting binary image are taken to be the shorelines. The Uniform Cost algorithm is used to find the least cost matches of the "scale-space images" of shorelines extracted from the Landsat image and those in the map giving priority to matches at higher levels of scale. A subset of those matches which are consistent are chosen to estimate the parameters of an affine transformation from the image to the map using a least squares method.