針對傳統參數活動輪廓模型存在對輪廓線初始位置敏感的缺點，提出了方向氣球力活動輪廓模型并應用于MRI 圖像分割。該模型利用底層圖像分割的結果確定外力的方向，使氣球力方向始終指向目標邊界，引導輪廓線變形。當輪廓線運動到目標邊界附近時，在高斯勢力作用下繼續變形，完成圖像高層分割。實驗結果表明,該模型與輪廓線初始位置無關，能實現MRI 圖像的自動分割。
關鍵詞: 圖像分割; 方向氣球力; 活動輪廓模型; 高斯勢力
Abstract: Traditional parametric active contour model is sensitive to the initial position. An improved external force for the active contour, called directional balloon force is proposed to address problem associated with initialization and is used to segment MRI images. In this model the direction of the force is decided by the results of low-level segmentation and always points to object boundary to make the contour deform. In the vicinity of object boundary, Gaussian potential force drives the contour towards boundary and high-level segmentation is implemented. The experiments of segmenting left ventricle MRI images show that this model be independent of the initial position and can segment image automatically.
Key words: image segmentation; directional balloon force; active contour model; Gaussian potential force