基于数学形态金字塔分解和流域分割的骨髓细胞图像分割

提出了一种基于数学形态金字塔分解和流域分割的骨髓细胞图像分割方法。根据形态金字塔的分解方法,将细胞图像经过形态金字塔分解后,先在低分辨率图像上进行流域分割,然后对低分辨率图像的分割结果进行复合得到原始图像的分割结果。实验结果表明,使用该方法能够获得较好的分割效果,并可提高分割算法的运算速度。

Faster Segmentation of Marrow Cell Images Based on Morphological Pyramid and Watershed Segmentation

Aim.In order to reduce the algorithm operation time,we propose a novel segmentation algorithm of marrow cell images based on the mathematical morphological pyramid and the watershed segmentation.We explain our novel algorithm in detail in the full paper;in this abstract,we just add some pertinent remarks to listing the three topics of explanation:(1) mathematical morphology;the subtopics are basic morphological operations(subtopic 1.1) and watershed segmentation(subtopic 1.2);(2) the decomposition of morphological pyramid;the subtopics are the decomposition of pyramid(subtopic 2.1) and morphological pyramid algorithm(subtopic 2.2);(3) morphological-pyramid-watershed-segmentation algorithm;the morphological pyramid is used in a scale space presentation,which is a multiresolution pyramidal structure created by the repetitive morphological filtering and subsampling of the original cell image;the cell image is watershed-segmented at a coarse level of the pyramid;the segmentation information is transferred to the finer levels and then transformed to construct the segmented results of the original cell image.Finally we performed experiments to test our novel algorithm.Fig.1 in the full paper shows a marrow cell gray image;Figs.2 through 6 give the results of processing it with our novel algorithm.The experiments show preliminarily that,as compared with existing algorithms,our novel morphological-pyramid-watershed-segmentation algorithm saves at lest 50% of computation time at no cost in segmentation quality of marrow cell image.