摘要
本文基于尺度空间理论提出了两种多门限化方法：第一种新方法先由精到粗求直方图在不同尺度下的表征，再由粗到精校正定位，最后由生命期长短来确定门限；第二种方法首先在一粗尺度上找特征结构，再由粗到精“聚焦”到最精尺度上获取门限。两方法克服了以往一般门限法的若干缺陷，不仅定位准确，而且普适性强。实验证明新方法是令人满意的。
Based on scale space theory, two methods for muti thresholding are put forth in this paper. One of them consists of three main steps: first, the scale space representations are derived from finer level to coarser level; then, the locations of structures are corrected from coarser level to finer level; finally, the thresholds are derived from the lifetime of structures. The other method finds the structures at a coarser scale, and then focus on finest scale step by step to derive the thresholds. Experiments show that new ways are effective and robust.
出处
《国防科技大学学报》
EI
CAS
CSCD
1999年第4期55-58,共4页
Journal of National University of Defense Technology
关键词
图像分割
尺度空间
多门限化
image segmentation, scale space, muti thresholding
