Object　detection　and　segmentation　is　an　important　direction　in　biological　image　processing.　Traditional　thresholding　and　labeling　methods　as　well　as　machine　learning　methods　are　the　two　predominant　ways　to　solve　this　problem.　In　this　article,　a　multi-threshold　algorithm　doing　cell　segmentation　is　developed　and　applied　on　Fluo-N2DH-GOWT1　dataset[1-2],　which　contains　low-contrast　and　noisy　biomedical　images.　Besides,　U-net　is　also　utilized　and　the　results　of　U-net　as　well　as　multi-threshold　algorithm　are　compared　to　better　illustrate　the　distinctions　of　the　two　algorithms.　©　2020　ACM.