Scan-in polygonal extraction of video images

A video image is extracted and stored as a geometric figure by generating a binary pixel map which contains, for each pixel in the image, one bit representative of the value of a characteristic feature of that pixel relative to a threshold range. A crack-following algorithm locates the border of a group of contiguous pixels all having the same binary value. This border defines the boundaries of the scanned image. It is smoothed by a recursive adaption technique to generate one or more polygons related to the shape of the image.