An edge enhancement technique for image segmentation based on resistive circuit simulation
T. Yoahida, Keiichi Hirano · 2002
The quality of image segmentation is often affected by that of edge profile extracted from a target image. The completely closed and continuous edge for a target object is preferable, but it is actually difficult to obtain such a perfect edge. This paper thus proposes a novel technique for broken edge linking and edge enhancement based on an idea of electronic-circuit simulation, and applies it to the target extraction problem. The proposed technique first maps the edge intensity into the value of the resistor in a circuit and calculates its current distribution. Then, the current in each resistor is mapped back to the pixel intensity. Due to the nature of the current flow, such a process can link broken edges as well as enhance weak edges. An application of this technique to a target extraction tool improves the quality of extracted objects. Several experimental results, which are given, demonstrate the effectiveness of the proposed approach.