摘要
提出将LEGION(LocallyExcitatoryGloballyInhibitatoryOscillatorNetwork)这种新的基于生物仿真结构的神经网络应用于分割心脏的超声图像 ,讨论了LEGION中参数对分割结果的影响。由于LEGION具有天然的并行、分布式处理的特性 ,与一些现有的分割方法进行了比较 ,结果表明本文所用的方法所得的分割图像边缘更为清晰具体 ,细节保留更为良好。
Image segmentation has always been one of the difficult problems in image processing. It is the key from image processing to image analysis and the first step of automatic image analysis. Ultrasound images of human heart, because of its noise and fuzzy boundary, are especially hard to segment. In this paper, a biologically simulated structure, LEGION (Locally Excitatory Globally Inhibition Oscillator Network), is first introduced and then applied to segment the sample images. Parameters of the algorithm are analyzed. And at last, the result is compared with those produced by some other methods.Due to its feature of parallel and distributed processing,images processed by LEGION generally have clearer boundary and better details than those segmented by some traditional methods.
出处
《四川大学学报（工程科学版）》
EI
CAS
CSCD
2001年第3期113-116,共4页
Journal of Sichuan University (Engineering Science Edition)
基金
国家自然科学基金资助项目! (6 96 310 2 0 )
