Abstract: Cells of microscopic cell images are often adhesive. In order to count, it is usually necessary to find the seed points of the adhesion cells accurately. This paper proposes an algorithm of counting the overlapped cells based on pixel block scanning with good adaptability. On the basis of determining the area of the maximum and minimum in connected sections, the length of the maximum and minimum rectangular pixel block with same width side is determined. And then, the two value image is scanned by the maximum pixel block with width of side automatic decreasing from left to right, from top to bottom. The area matched to a near-circular pixel block is replaced with a seed point to get a seed point image and counting. A lot of experiments show that the method proposed in this paper has a significant effect on counting overlapped cells with different areas and has good adaptability.

Key words: Microscopic cell image, cell counting, rectangular pixel block, near-circular pixel block, overlapped cell

摘要： 对于显微细胞图像中的粘连细胞，为了计数通常需要比较准确的找出粘连细胞的种子点，提出了一种适应性较好的像素块遍历的计数算法用于对粘连细胞的计数。在确定最大与最小连通域的像素面积的基础上，算出最大与最小长宽相等的矩形像素块边长，并用最大像素块对二值化图像进行递减遍历。将与像素块匹配的地方近圆像素块替换，得到种子点图像并计数。实验结果表明，特别适用于不同大小的细的各种粘连程度的粘连细胞计数，效果显著且适应性好。

关键词:

"> 显微细胞图像, 细胞计数, 矩形像素块, 近圆像素块, 粘连细胞