Abstract: A method for spine segmentation based on adaptive mathematical morphology was proposed in order to increase the correct rate of Pyrrophyta recognition. At first, the pixel-width was introduced, and the optimal structure element was computed by pixel-width histogram and area distribution automatically, and then the spine was extracted by mixed mathematical morphology operations. Finally, two kinds of local biological morphology feature parameters were constructed and their visual invariance was proved. The experiments results show that the optimal structure element can be computed by different Pyrrophyta cell and this method is of high precision and fast speed.

Key words: Pyrrophyta, Spine segmentation, Adaptive structure clement