This thesis describes a flexible and efficient pre-processing system which extracts the set of graphical output primitives from a line drawing. Since the system does not use any internal model for the extraction task, the input can be adopted from a wide variety of line drawings. The system follows the general sequence of processing steps for data reduction, image scanning, text separation, thinning, approximation, and encoding of lines. But each step is implemented by the use of various improvement strategies upon the existing well-known algorithms to increase the efficiency of such system. By the stepwise improvements, the pre-processing system runs fast and generates a rather compact description. In addition, there is a high degree of data reduction.
This system has been implemented in a "C" programming language on a Trigem/AT machine running under a Xensix System V operating system. Five line drawings are used to analyze the performance of the system.