Extraction of outline in arbitrary shape from binary images using genetic algorithm.

This paper proposes a method of extracting the outline of an arbitrary shape in an image by using a genetic algorithm (GA). The image under consideration is a binary image containing a disconnected outline and noise. The proposed method extracts the outline from such a binary image as visually recognized by a human observer. First, the single outline extraction method is proposed, which extracts a single outline from the image. In the proposed method, Delaunay triangulation is applied to the point set in the binary image, and the graph containing the outline is derived. Then, the points and edges which can be recognized locally as the noise are eliminated. The outline is extracted from the graph after noise elimination using the GA. In the GA, gene coding is applied in which the polygons in the graph correspond to genes. It is shown experimentally that the outline of an arbitrary shape can be extracted by the proposed method. Next, the single outline extraction method is extended to a method by which multiple outlines contained in an image can be extracted. An experimental result is presented in which multiple outlines are extracted from an image. © 2005 Wiley Periodicals, Inc. Electron Comm Jpn Pt 2, 88(2): 32–46, 2005; Published online in Wiley InterScience (<URL>www.interscience.wiley.com</URL>). DOI 10.1002/ecjb.20146 [ABSTRACT FROM AUTHOR]