Abstract: An algorithmically based method and apparatus for registering halftone color separations using edge based registration techniques and three registration points on each film separation in order to handle image stretching in both directions. Improved techniques include the use of gray level transformation, an improved U.S.C. predicate, double accuracy interpretation after systematic convolution sampling, finer angle sweeping to achieve smoother edge transitions around sharp corners, elimination of gaps in the edge continuity, enhancement of final edge linking, a more accurate and robust segment matching algorithm, invocation of a registration quality index for quality control checking of each segment picture registration, and reduction of machine cycle time by performing computations concurrently with machine movement.

Abstract: An algorithmic based method and apparatus for registering halftone color separations using edge based registration techniques. Digital pictures are taken at two locations on each film, and, following optimal thresholding of the data, the edges are located by application of a Laplacian of a Gaussian edge detection operator. Sampling is used to reduce the computational requirements, and full resolution is restored by interpolation. The edge locations are corrected for operator edge bias, and the corrected edge points are linked into segments which are registered by the parameter space method. The initial offset is corrected for size differences by application of least squares correction techniques.