A new approach for intelligent object picking in line drawing images

This paper proposes a new approach for picking raster entities in the intelligent raster editing process of line drawing images. The proposed method is based on the run-graph encoding principle and involves several key steps including: tracing the line in both positive and negative directions, recognizing line junction areas, property consistent tracing, and reconstructing line junction areas. It is shown, both in theory and experiments, that the algorithm provides a robust performance in recognizing the line junction areas and their reconstruction. By this approach, two kinds of basic raster entities could be picked up completely even in noise disrupted line drawing images.