以區域生長與形態學方式來估計大鼠聲門面積之研究

Abstract

聲門動態變化在喉生理學研究不論是在人類聲門或動物聲門上都是相 當重要的課題。本研究是採用大鼠聲門來進行研究，且以往有關大鼠聲門文獻在研究方法上都有計算速度太慢或準確性不足的缺點，故本研究提出一影像處理方法來改善上述缺點。其中，影像處理方法來計算聲門面積可分為以下三個處理步驟：

 擷取紅色(R)成分影像。

 利用區域生長(Region-Growing)來界定對象或聲帶的開放區域，並獲得該地區的聲門。

 應用形態學(Morphology)來刪除孤立對象的區域。

實驗證明，擷取R成分影像後，經由區域生長能夠有效的分割出大鼠聲門區域，接著運用形態學方式來去除雜訊及小的結構區域，使得準確地計算出大鼠聲門面積。最後在與Otsu's method及active-contour method做比較，並從實驗結果證明本研究方法在大鼠聲門影像分析上有較好的準確率並且能夠加快影像處理速度。

The dynamic change of glottal area is a critical subject in the filed of laryngeal physiology. However, efficient methods for estimating glottal area were discussed rarely in previous literatures. In this regards, a glottal area estimation method by combining region growing technique and morphology operator was proposed in this dissertation. The proposed approach involves three major process steps including 1. Extracting red color components from the target image; 2. Using region-growing to delineate the vocal fold opening region; 3. Applying a morphology operator to remove the isolated subject regions and to obtain the area of the glottis. Several experimental results demonstrate both the accuracy and computational efficiency for the glottal area estimation can be greatly improved by using the proposed method.