Vectorisation

Abstract

The final step in the raster to vector transformation is to move from a skeletonised or contoured binary image to the set of primarily geometrical primitives that make up the vector descriptions outlined in Chapter 2. As always, the detailed properties required of a vector description will depend upon the type of drawing being processed, the planned sequence of subsequent operations and the overall goals of the application. However, possibly because vector descriptions are closer to those used by humans in their interpretation of line drawings than the pixel-based representations considered previously, human observers are generally skilled in evaluating the quality of vector output. If a vectorisation system is to be accepted into everyday use the geometrical primitives it presents to the user must be of the right type (straight lines must be described by straight lines, circles by circles, etc.), fit the underlying skeleton or contour data accurately and be produced reasonably quickly. A fairly standard trade-off exists between these last two aims.