摘要
为提高灰度图像分割的效果，提出了一种新的基于图像间模糊散度的阈值化算法及它在多阈值选择中的推广算法。算法采用模糊集合分别表达分割前后的图像，通过最小模糊散度准则实现图像分割中最优阈值的自动提取。算法针对图像阈值化分割的要求构造了一种新的模糊隶属度函数，克服了传统Ｓ－函数带宽对分割效果的影响。将其与多种经典的阈值化分割算法一起，对不同类型的测试图像进行分割比较的结果表明，新算法有很好的通用性和有效性。
A new thresholding algorithm and its multi threshold extension are presented to improve the performance of image segmentation. The algorithms are based on computing the fuzzy divergence between original image and its segmented version. The optimal threshold is searched through a minimum fuzzy divergence criterion. A new fuzzy membership function is defined in the light of requirements of image thresholding which can overcome the influence on segmentation caused by the classical S function. Compared with several traditional thresholding methods by applying them to various test images, the effectiveness and generality of our new algorithms are shown.
出处
《清华大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
1999年第1期47-50,共4页
Journal of Tsinghua University(Science and Technology)
基金
国家自然科学基金
清华大学智能技术与系统国家重点实验室部分资助
关键词
图像阈值化分割
模糊集合
图像分割
模糊散度
image thresholding
fuzzy set
fuzzy membership function
fuzzy divergence
