Abstract: Aiming at the burr phenomenon and skeleton redundancy in current image skeleton extraction, a method based on flood-fill is proposed. The target image is pre-processed to obtain a threshold binary image and the foreground and background are segmented. A mask image and axis map are constructed to provide data sources and judgment basis for the subsequent calculation process. The diffusing water process is simulated by traversing the horizontal grayscale histogram. The foreground contour of the mask image is shrinked. According to the change of the connected components of the neighborhood, a skeleton feature point is determined . When the contour shrinking is finished, the image skeleton is obtained. The experimental results show that the method can effectively extract the image skeleton, the burr phenomenon and redundant structure are significantly reduced, and the execution speed is fast.

Key words: skeleton extraction, flood-fill, mask image, axes image, water surface confluence