局域化互信息度量的ACM下医学图像的分割

Medical image segmentation based on ACM of localized mutual information

大量医学图像中存在灰度不均匀现象使传统方法很难获得理想的分割结果,针对此问题,将图像的局部统计特征引入互信息度量的分割模型中,考虑不同组织间的方差差异,提出一种局域化互信息度量的活动轮廓模型(ACM),以提高灰度不均匀情况下目标边界识别的精确度。此外,采用一种无需重新初始化的水平集函数规则化方法,演化稳定,收敛速度快。最后,以医学图像分割实验对算法进行了验证,对比实验表明分割和偏移场矫正结果都更精确。

Due to the inhomogeneous intensity in most medical images, it is usually difficult for the traditional methods to obtain desired segmentation results. Aiming at this problem, by using the local statistical characteristic of an image, an active contour model based on local mutual information, which considers the variance differences among different tissues, is presented to improve the target boundary identification accuracy in the case of inhomogeneous intensity. And the level set function without re-initialization is regularized, its evolution is stable with a faster convergence rate. The proposed method was applied to segment medical image with promising results. The comparative experiment demonstrates that both the segmentation and bias correction results are more accurate.