基于最小生成树及形状的细胞图像分割

CELL IMAGE SEGMENTATION BASED ON MST AND SHAPES

为了能够对细胞图像数据进行准确的分析,针对人体医学细胞图像的特点,作者应用图论思想提出了一种基于细胞形状的分割算法。该算法先对图像进行平滑、锐化等预处理,然后用图论的最小生成树(minimum spanning tree,MST)算法对图像中粘连细胞进行初步的分割,再结合细胞边界的曲率进行再分割。通过一系列细胞图像的试验,该算法能够分割一般的医学细胞图像,与传统的图论分割算法及分水岭算法相比,该算法对有细胞粘连的图像具有较好的分割效果。

To be able to analyse cell image data accurately, the paper proposes a segmentation algorithm based on graph theory and cell shapes. The algorithm includes image smoothing, sharpening and other pre - treatments in image pre - processing. After that, it ap- plies graph theory MST algorithm to segment the image, and finally re - segmentation them combining with the curvature of the cell boundary accurately. By experiments, this algorithm can segment normal cell images. With comparing to the traditional graph segmen- tation algorithm and watershed algorithm, the studied algorithm is better for the image segmentation of touching cells.