摘要
为实现数字图像边缘的有效检测与提取,借助BP神经网络,采用了改进的最速梯度下降法,通过对动量项的合理选择,有效地实现算法的快速收敛。为提高算法的执行效率,采用直接编程和对图像采用分块的思想,并给出了算法实现的方法和步骤。用Matlab软件对灰度图像进行了仿真,并将仿真结果和传统的方法进行了比较,结果表明,所设计的网络边缘检测优于传统方法,并具有较好的泛化能力。
To realize the edge detection of digital image, an improved deepest decedent method is employed through BP neural network. This method realize the rapid convergence of the algorithm by properly selecting item of momenturn. At the same time, to improve the efficiency of the algorithm, directly program and blocked image are adopted. The way and steps of the algorithm are also given. In the end, the simulation using Matlab is performed, compared with several traditional edge detections, this method is prior to those and the method has generalization ability.
出处
《计算机工程与设计》
CSCD
北大核心
2011年第6期2146-2149,共4页
Computer Engineering and Design
基金
河南省教育厅自然科学研究基金项目(2009B510007)
关键词
BP神经网络
图像边缘检测
最速梯度下降
分块
泛化能力
BP neural network
image edge detection
deepest decedent method
image blocks
generalization ability
