摘要
本文提出采用多层前向网及改进的ＢＰ算法，对经Ｃａｎｎｙ最佳边缘检测算子处理噪声图象后得到的高、低门限图进行不连续轮廓的并行“修补”，能够较快地得到信噪比高的边缘图象．
This paper describes a method to detect noise-blurred image's edge by feedforward neural network with modified BP algorithm. Owing to neural network's parallel characteristic, non-continuous contour may be marked by low-thresholded image and high-thresholded image, both of which were obtained from Canny' s optimal edge operator, more quickly and precisely.
出处
《上海大学学报（自然科学版）》
CAS
CSCD
1996年第5期549-555,共7页
Journal of Shanghai University:Natural Science Edition
