以Fuzzy C-Means色彩量化法應用於細胞計數

生物檢測的需求隨著人類對健康的重視而更為重要，檢測時以人工細胞計數費力又耗時，且需要依賴專業人員的能力，專業人員的生理與心理狀態常會影響細胞觀測結果的準確性。為了解決以顯微鏡人工計數的不確定因素，因此本論文提出一個快速與準確的自動化計數工具。本論文利用色彩量化演算法與標記演算法來進行細胞計數工作，相較於使用邊緣偵測及相關去除雜訊演算法的研究方式，本論文所提出的方法只需將樣本作色彩的量化分組後，針對量化的結果進行標記演算法的運算即可，因而可以達到快速且準確的結果。　本論文實現的計算細胞方式，可依照細胞圖片的特性作相關的量化條件調整，經過調整後，其準確率約為95%，可以有效降低人力成本的支出及人為因素的誤差率。

關鍵字

細胞計數；色彩量化；標記

並列摘要

The demand for biological testing becomes more and more important as humans pay more attention to their health. Conventional manual cell counting process in biological testing is laborious and rather time consuming. Besides, the accuracy of counting results relies on the ability of professional operators. The physical and psychological conditions of the operators often affect the accuracy of the observation results. Accordingly, we propose in this thesis an automated counting tool with a fast and accurate way to solve the uncertainty under the microscope observation, and hence to reduce the possible human errors. This thesis uses color quantization algorithms and fast connected-component labeling (FCCL) for counting cell, rather than the familiar methods using edge detections and noise removal algorithms. By tuning the condition of color quantization for cell counting method according to the cell image properties, we can achieve the counting accuracy rate of 95%, and efficiently reduce the labor cost and the error rate caused by human.