摘要
MR图像中常含有偏移场以及噪声现象,传统的高斯混合模型无法得到正确的分类.在高斯混合模型目标函数中加入偏移场估计与噪声去除,完善其分类效果,使分类结果较好地克服偏移场和噪声影响.实验表明,该算法在得到较准确的分类结果的同时还能很好地估计偏移场.
The conventional GMM mode can＇t get the precise classification results of magnetic resonance images, which are disturbed by bias field and noises. In order to overcome these limitations, bias field estimation and noise removal are incorporated in GMM mode in this paper. The new mode can reduce the effect of bias field and noise so as to get better classification results. Experiments on the segmentation of magnetic resonance images show that this mode has better effect in image segmentation and is able to have the bias field well estimated at the same time.
出处
《南京信息工程大学学报（自然科学版）》
CAS
2009年第3期208-212,共5页
Journal of Nanjing University of Information Science & Technology（Natural Science Edition）
基金
国家自然科学基金(60973157)
江苏省高校自然科学基金(08KJB520004)
关键词
高斯混合模型
偏移场
去噪
Gaussian mixed mode
bias field
noise removal
