基于多进制小波变换的渐进式纹理图像检索

¹ 中国科学院遥感应用研究所遥感科学国家重点实验室,北京市大屯路甲3号,100101;² 武汉大学测绘遥感信息工程国家重点实验室,武汉市珞喻路129号,430079

基金项目: 国家973计划资助项目(G2000077906)

详细信息

Progressive Texture Image Retrieval Based on M-Band Wavelet Features

¹ State Key Laboratory of Remote Sensing Information Sciences, IRSA, CAS, 3 Datun Road, Beijing 100101,China;² State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, 129 Luoyu Road, Wuhan 430079, China;

Funds: 国家973计划资助项目(G2000077906)

摘要: 提出了一种基于多进制小波变换的渐进式纹理图像检索方法,纹理图像的特征通过多进制小波分解结果来描述,用低频子图小波系数标准方差和多进制小波直方图结合相应的相似性距离函数,实现目标图像数据库由粗到精的渐进式检索。通过Bordatz和USC纹理图像数据库来检验本方法的精度和效率,获得了较理想的试验结果。
- 多进制小波 /
- 渐进式图像检索 /
- 小波系数标准方差 /
- 小波直方图 /
- 纹理
Abstract: A M-band-wavelet-based features for progressive texture image retrieval method is presented. The information of a texture image is acquired through M-band wavelet decomposition of the original image. The integration of two feature descriptors, wavelet-coefficient standard deviation of the approximation sub-image and M-band wavelet histogram, and their corresponding distance similarity measure functions are used to realize progressive retrieval of target texture image databases. Finally, a texture database obtained from Bordatz album and USC image database is used to check the retrieval performance of the proposed method.
- M-band wavelet /
- progressive image retrieval /
- wavelet-coefficient standard deviation /
- wavelet histogram /
- texture