摘要
提出一种具有近似旋转不变性的改进Gabor小波变换纹理特征提取方法,由小波变换系数模的均值和标准方差组成特征向量表示图像内容,利用10幅B rodatz纹理图像经过旋转、分割组成的图像数据库进行了检索测试,并与传统Gabor变换和二元树复小波变换特征提取方法的分类结果进行了比较分析,实验表明本文方法有效地提高了图像检索精度.
In this paper, a method of improved images texture feature extraction based Gabor wavelet transform with approximated rotation invarianee is proposed, the image contents were denoted by feature vector generated by computing mean and standard deviation of modulus coefficients of Gabor wavelet transform, retrieval experiment were carried on images database by rotating and splitting 10 Brodatz texture images,as well as comparison and analysis were given with traditional Gabor transform and dual - tree complex wavelets transform, results of experiments indicate that the presented approach provide superior images retrieval precision.
出处
《云南民族大学学报（自然科学版）》
CAS
2009年第4期287-291,共5页
Journal of Yunnan Minzu University:Natural Sciences Edition
基金
国家自然科学基金重大项目(50490270)
国家杰出青年科学基金项目(50225414)
关键词
GABOR小波变换
二元树复小波变换
特征提取
纹理分类
旋转不变性
Gabor wavelet transform
dual - tree complex wavelets transform （ DT - CWT）
feature extraction
texture classification
rotation invariance
