摘要
纹理为多类图象分析提供了重要的特征。本文提出一种新的多分辨率纹理分类方法，该方法采用称为小波帧的冗余小波分解，从而获得具有稳定性和平移不变性的特征描述。同时，为适应纹理的准周期特性，我们采用树型变换包分解策略，即根据某种分解准则，同时对低频信道及高频信道进行分解。另外，我们在算法中还采用自适应的分解方法，一旦给定条件满足，分解终止，整个分类过程结束，从而避免了不必要的分解与计算，提高了计算效率。实验表明，本文提出的方法无论在分类性能和计算效率上都优于现有的分类方法。
A new multi resolution texture classification method is presented in this paper. The method adopts redundant wavelet, which is called wavelet frame, to decompose and then to achieve the characteristic description of stability and translational constancy. Meanwhile, in order to fit the quasi periodic characteristics of texture, the tree transform packet′s decomposition strategy is applied to split low and high frequency signals. Adaptive decomposition is implemented to avoid unnecessary decomposition and computation. Simulation results show that the proposed method has better performance in classification and in computation efficiency than the current classification methods.
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
1997年第6期520-526,共7页
Journal of Computer-Aided Design & Computer Graphics
关键词
纹理分类
小波帧
小波包
图象分析
texture segmentation, wavelet frame, wavelet packet.
