摘要
Brushlet是一种新的图像方向信息分析工具,其能量特征已被应用于纹理分割、分类以及去噪等领域。该文利用Brushlet变换为复函数这一特性,提取其能量及相位信息作为纹理分类特征。通过对Brodatz纹理图像库中均匀、非均匀以及全部图像进行分类实验,较之单一能量特征的分类方法,Brushlet复特征取得了更好的分类性能。
Brushlet is a novel tool for image orientation analysis, whose energy feature is adopted in texture segmentation, image classification and denoising. In this paper, the property of Brushlet is used: the transform is a complex value function with a phase, the energy and phase information are adopted as a fused feature for texture classification. Experiments on homogeneous, inhomogeneous images and total Brodatz texture alblum prove that the complex feature of Brushlet outperforms the method based on single energy.
出处
《电子与信息学报》
EI
CSCD
北大核心
2007年第10期2301-2304,共4页
Journal of Electronics & Information Technology
基金
国家自然科学基金(60505010
60472084)
国家"973"重点基础研究发展规划项目基金(2001CB309403)资助课题