基于自适应分割与偏移场估计的活动轮廓模型

针对传统活动轮廓模型无法快速、准确、强鲁棒性地分割灰度不均匀图像的问题,提出了偏移场估计与图像分割相结合的新型混合活动轮廓模型。首先,通过对图像进行模糊聚类分析,提出带有模糊隶属度函数的新型偏移场估计模型,提高了模型对图像灰度信息的估计与提取能力。其次,利用图像信息熵构造了自适应尺度算子(adaptive scaling operator, ASO),改善了模型的分割效率及对初始轮廓和噪声的鲁棒性。最后,通过将偏移场估计模型和ASO融入到能量泛函中,提出新型混合活动轮廓模型。实验结果表明,该模型不但对初始轮廓和不同种类噪声具有较强的鲁棒性,而且对不同程度的灰度不均匀图像具有较高的分割准确度与分割效率。

Active contour model based on adaptive segmentation and bias field estimation

In order to solve the issue that traditional active contour models cannot quickly, accurately and robustly segment inhomogeneous intensity images, a hybrid active contour model combining bias field estimation and image segmentation is proposed. Firstly, through fuzzy clustering analysis for images, a bias field estimation model with the fuzzy membership function is proposed, which improves the ability to estimate and extract image intensity. Secondly, an adaptive scaling operator (ASO) is defined based on image information entropy, which improves segmentation efficiency and robustness to initialization and to noise. Finally, a hybrid active contour model is proposed by incorporating the bias field estimation model and the ASO into an energy functional. The final experiment results show that the proposed method not only has strong robustness to initialization and noise, but also has higher segmentation accuracy and segmentation efficiency for different degrees of inhomogeneous intensity images.