摘要
提出一种基于最小错误率和快速水平集的图像分割方法,通过对速度项和停止条件的重新设计,实现了快速而有效的图像分割。算法采用模式分类思想,以统计直方图来近似目标和背景区域的概率密度,对基于最小错误率的判别函数进行平滑滤波以获得外部速度,从而实现曲线的进化;同时,分割过程在分类错误率达到最小时停止。实验结果表明,本文算法对弱边缘、低对比度灰度图像具有较好的分割效果,且具有较强的抗噪性能;在分割速度上,本文算法也明显优于几种已有算法。
An image segmentation is proposed based on fast level set and least error rate.It designs new evolution velocity item and stop condition to realize the fast and effective image segmentation.The algorithm adopts the idea of pattern classification,approximates the probability density of object and background by statistical histogram,utilizes a filter window to smooth discriminating function and gets external evolution velocity to realize curve evolution.Meanwhile,the segmentation process is stopped when the error rate is the least.Experiments show that this algorithm can effectively segment the weak edge and low contrast intensity image,and it has strong anti-noise capability,further more,it is much faster than several existing algorithms.
出处
《光电子．激光》
EI
CAS
CSCD
北大核心
2010年第1期130-135,共6页
Journal of Optoelectronics·Laser
基金
国家高技术研究发展计划(863)资助项目(2007AA701206)
关键词
概率密度
水平集
最小错误率
图像分割
probability density
level set
least error rate
image segmentation
