摘要
边缘检测是图像分析识别必不可少的环节,是一种重要的图像预处理技术。虽然传统的算子算法对边缘的检测速度快,但其得到的往往是断续的,不完整的边缘信息,且这类检测方法对噪声比较敏感,在检测噪声污染图像时会得到许多虚假的边缘。利用CP神经网络对灰度图像的边缘进行检测,但考虑到神经网络训练量过大的问题,先利用传统算子对图像进行边缘处理,将处理后的图像做为神经网络的输入。实验结果表明,该方法得到的边缘图像边界封闭性好,具有较好的抗噪特点。
Edge detection is an essential step in image analysis and recognition, and it is an important technology in the image preprocessing procedure. Although the traditional operator＇s detection speed is fast in edge detection, its edge information is discontinuous. Traditional detect method is sensitive to noise and easily gets many false edges when detecting the noise image. An edge detection method of gray-scale image based on CP neural network is proposed in this paper. Considering the problems of too many trained samples, the traditional algorithm is used first, and then processing results are the input of the neural network. Experiment proves that the enclose function for edge is good and with better anti-noise characteristic when using this method.
出处
《山西电子技术》
2006年第1期52-54,共3页
Shanxi Electronic Technology
