摘要
提出了基于模糊聚类原理的图像统计特征识别技术。通过分析象素灰度矩阵信息,提取出图像独立的统计特征量,建立起相应的图像统计特征模型空间Ω。由此,计算出各图像类的模糊相似系数矩阵,再运用聚类分析的传递闭包法将其改造成为模糊等价矩阵,划分出图像等价类,进而实现对目标图像的识别。实验结果表明,该方法能获得很好的图像识别效果。
Based on fuzzy cluster analysis, a scheme for recognizing image is presented. To analyze the spatial knowledge of pixels' hue matrix, and extract statistic features of image, a fuzzy similitude matrix is defined to describe the images' feature space 'Ω'. Introducing the fuzzy cluster analysis, a corresponding fuzzy equivalence matrix is obtained over the fuzzy similitude matrix by computing the transferred package. Under the biggest threshold 'λ', images are classified to two groups. The classification illustrates which group the unrecognized image belongs to, and what the unrecognized object is. A prototype system was developed to evaluate the effectiveness of the approach.The experiment shows that the proposed approach is effective on image recognition.
出处
《四川大学学报(工程科学版)》
EI
CAS
CSCD
2003年第3期83-86,共4页
Journal of Sichuan University (Engineering Science Edition)
关键词
聚类分析
特征提取
图像识别
相似关系
cluster analysis
feature extraction
image recognition
similitude relation