The clustering phenomenon often appears in the cell image auto reading system

i.e.

some cells overlap or touch together to form a big area. It is necessary to design an effective algorithm to separate the clustering cells into single ones. Most of the existed separating algorithms are based on two hypotheses: 1) at points where cells touch

the cluster boundary tends to form an acute angle; 2) at points where the cells touch

the optical density is relatively low

otherwise it will be very difficult to separate them properly. But in the actual cell images

the two hypotheses are very difficult to be met. To solve this problem

an algorithm based on boundary stripped is presented in the paper. The boundary of clustering cells is stripped layer by layer in the algorithm. During stripping

judgement is made to determine whether the splitting has happened

and then the actual separation is taken in the original cell image. It does not require the clustering cells should necessarily meet the two hypotheses. The experiment result shows that the algorithm is effective.