|
Cell segmentation is an essential step for cell observation and analysis at individual level and group level. However, in high throughput images, touching cells detection and separation remains a challenge. In this study, a touching cells segmentation method for negative phase contrast images is proposed, which takes advantage of both intensity information and shape information. An adaptive binarization is adopted in order to obtain the preliminary segmentation and the shape information of cell regions. Then the intensity information and the shape information are fused by means of gray weighted distance transform, and the transformation result is utilized to detect suspected touching cells. A region skeleton based touching cell separation method is implemented to split the actual touching cells and remain the actual individual one using the fused information. Experiments are conducted on comparison with other touching cell separation methods and on whole cell image segmentation, the result of which shows an improved performance of the proposed method. |
|
Keywords:pattern recognition and intelligent system; touching cell separation; gray weighted distance transform; region skeleton; cell shape |
|