用BP网络进行彩色图像分割和边缘检测  

Segmenting the Color Image and Detecting the Edge by BP NN Method

作　　者：熊联欢[1] 胡汉平[1] 李德华[1] 李泽宇[1] 

机构地区：[1]华中理工大学图像识别与人工智能研究所图像信息与智能控制国家教育部开放实验室

出　　处：《华中理工大学学报》1999年第2期87-89,共3页Journal of Huazhong University of Science and Technology

基　　金：国家自然科学基金

摘　　要：采用ＢＰ网络模型，研究了彩色图像分割和边缘检测的神经网络方法．选取训练样本图像，并分别以区域增长分割法和Ｓｏｂｅｌ边缘检测方法所得结果为导师信号，将图像的特征向量采用ＢＰ算法进行训练，然后对实测图像进行分割和边缘检测．与采用区域增长法所得分割结果和采用Ｓｏｂｅｌ边缘检测方法所得边缘检测结果进行比较。The neural network method for segmenting the target and detecting the edge from the color image is studied. The adapted network model is BP network. The segmented images by the region growing method or the edge images detected by the Sobel edge detecting method are used as the training set, and the BP algorithm is used to train the feature vector of the image. The segmentation and the edge detecting of the testing image can be processed. The experimental results show that the segmentation and the edge detecting effects of the BP network method are as good as the region growing method and the Sobel edge detecting method.

关 键 词：彩色图像分割 边缘检测 BP网络 图像分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]