A new Automatic Seeded Region Growing Algorithm

Generally speaking, the object well dealt by region growing algorithm does not have clear boundary or gradient regularization. Therefore, the explorer method is proposed in this paper, which makes it possible to obtain a seed region following the rapidly self-adapting multi-threshold segmentation. Each boundary pixel is regarded as an encampment, which can send four explorers to its four neighbors. In the particular case of coordination between the encampments, every explorer with its limited supplies, according to the rule, can supplement its consumption and then overcome the block created by the differential value of grayness between the pixels. In the course of iteration, the segmentation of images can be fulfilled. The measurement and test of a series of standard images shows that the method is practicable and effective in the segmentation of an object image.