摘要
本文对Mallat基于小波变换的边缘检测算法作了改进,使得在不同尺度下的边缘移位减小到1个像素以内.本文提出了一种基于小波变换的多尺度边检算法.与传统的Guass尺度空间算法相比,本算法具有定位准确计算量小的优点,并可利用多尺度空间信息,识别边缘的特性.作为一个例子,本文给出了一个实验.
A novel multi-scale filtering based on wavelet transform is proposed in this paper. Compared with traditional Guass scale space, wavelet scale space has some important merits: low computational complexity and high location quality. Using multi-scale information, we can estimate the edge singularity and how smooth the edge is. One can obtain different kinds of edges according to different destinations.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
1996年第3期251-257,共7页
Pattern Recognition and Artificial Intelligence
基金
国家攀登计划认知科学(神经网络)重大关键项目重点课题的资助
