面向脑MR影像分割与偏置场修正的FCM方法

传统模糊C-均值(fuzzy C-means,FCM)算法以及结合空间信息的改进方法在分割脑磁共振(magnetic resonance,MR)影像时对噪声十分敏感,且无法消除脑MR影像中的偏置场。针对上述问题,提出一种面向脑MR影像分割与偏置场修正的FCM方法。该方法充分利用图像中的空间局部信息和非局部信息,分别构造出多局部信息模糊因子和非局部权重,在提高算法抗噪性的同时,尽可能多地保持图像细节;建立偏置场模型,以去除脑MR影像中的灰度不均匀性;将提出的多局部信息模糊因子和非局部权重嵌入到带有偏置场模型的FCM方法中,以实现噪声和偏置场条件下的脑MR影像分割。实验结果表明,与其他方法相比,所提方法对脑MR影像中的噪声具有更强的抑制能力,且能够有效消除偏置场对脑MR影像分割的影响,具有更优的分割性能。

Fuzzy C-means method for brain magnetic resonance image segmentation and bias field correction

As traditional fuzzy C-means (FCM) algorithm and its extensions combined with spatial information are quite sensitive to noise and incapable of removing bias field when segmenting brain magnetic resonance (MR) image. A new FCM method is proposed for brain MR image segmentation and bias field correction. To deeply dig out the spatial local and nonlocal information in an image, a novel multiple local information fuzzy factor and non-local weight are respectively presented to enhance noise robustness and preserve more image details. Then the bias field model is constructed to overcome the effect of intensity inhomogeneity. Finally, the new multiple local information fuzzy factor as well as the non-local weight is embedded into FCM framework with bias field model for brain MR image segmentation. Experimental results show that the proposed scheme is more robust to image noise compared with other methods, and it can effectively reduce the impact of bias field on brain MR image segmentation, thus achieving better segmentation performance.