摘要
论文通过分析最大信息熵双阈值分割算法原理,根据先验知识,设置目标出现最小最大频度条件,改变自适应双阈值计算条件,加快计算速度.同时通过计算分割后各个目标相关特征,利用BP神经网络技术对每个目标进行评价,根据评价结果修改原输入图像相应目标区域,并引入多层次迭代机制,实现了图像的多层次综合分割.实验结果表明,基于分割评价的多层次自适应双阈值分割算法提高了算法的通用性和分割的精确性,但是算法复杂性增加.
In this paper,a multi-level and adaptive dual-threshold segmentation algorithm is presented.According to the transcendental knowledge, the method quicken the most information entropy dual-threshold segmentation algorithm by setting the least and most frequency of target. At the same time, each segmented target was evaluated by caculateing relative characters and using BP network. Then according to the result,multi-level segmentation was implemented by amending target area and introduing multi-level iteration mechanism. The experiment result show that the algorithm greatly improve the effect of the image segmentation and make algorithm more all-purpose strengthen and more aUurate,but the algorithm complexity increases.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2009年第4期750-752,763,共4页
Acta Electronica Sinica
关键词
自适应双阈值分割
分割评价
多层次迭代
BP网络
adaptive dual threshold segmentation
segmentation evaluation
multi-level iteration
BP network
