中文摘要：

文章提出了一种新的脑磁共振图像自动阈值分割算法，该算法综合多分辨率和多上下文的思想．具有更好的抗噪声能力。多分辨率阚值法只考虑了多尺度下灰度级的一致性，不能兼顾图像各局部之间的不均匀性．而多上下文方法恰好弥补了这方面的不足。临床脑磁共振图像实验的结果表明。多分辨率多上下文的算法（MRMC）改进了分类精度，抗噪声能力优于单上下文的方法。

英文摘要：

This paper presents a new threshold algorithm for automatic sesmentation of brain magnetic resonance images. In addition to the muhiresolution framework, the algorithm applies muiticontext models to jointly use their distinct advantages, and has better resistance to noise. The multiresohtion threshold method only considers the glay- scale level of consistency, and does not able to attend the local heterogeneity in images. The multicontext method just makes up the shortage. Experimental results on clinical brain MR images show that the multiresolution and multicontext algorithm（MRMC） improved classification accuracy over the single context method.