摘要
通过提取原图像的同现矩阵和小波包分解后系数矩阵的统计量,构成了声纳图像的纹理特征集。通过实验比较进一步证明了该特征集作为纹理表示的有效性。
The texture feature set of sonar images is established by extracting co-occurrence matrix and statistical features of wavelet package coefficients of images. The results of experiments further prove the efficiency and effectiveness of this method as the represent of image texture.
出处
《计算机工程》
CAS
CSCD
北大核心
2004年第14期113-115,共3页
Computer Engineering
关键词
声纳图像识别
同现矩阵
小波包变换
特征维数
Sonar image recognition
Co-occurrence matrix
Wavelet package transform
Feature dimension
