多尺度H参数纹理特征的子带算法  

Wavelet Transform Based Multiscale Hurst Parameter Texture Features and Its Application

作　　者：李艳[1] 彭嘉雄[1] 

机构地区：[1]华中理工大学图像识别与人工智能研究所图像信息处理与智能控制教育部重点实验室

出　　处：《电子学报》2002年第7期1041-1043,共3页Acta Electronica Sinica

基　　金：北京大学视觉与听觉实验室开放基金资助课题 (No .0 9)

摘　　要：扩展的自相似模型 (ESS)是一种广义的分数布朗运动模型 (fBm) ,它的多尺度H参数与粗糙度之间是对应的 ,因为不要求粗糙度的尺度不变性 ,所以能够区分大多数自然纹理 .它的结构函数计算是基于图像在一定尺度上的灰度差 ,这可以用小波变换低频分量的一阶差分去定义 .由于小波变换具有抑制噪声的能力 ,由此导出的特征具有更好的抗噪性能 .实验证明对卫星遥感图像达到了较高的分类正确率 .同时也说明 ,纹理的自相似特性在低频分量上的体现更突出 .The extended self-similar model (ESS) is a general fractional Brownian motion (fBm) model. Its multiscale Hurst parameters have relations with roughness. In the meantime, it doesn't require the roughness to be scale-invariant as fractal dimensions do. The multiscale Hurst parameters can discriminate a large number of natural textures and are suitable to be used for texture classification. Its structure function is based on the difference of the gray levels at some scale. This can be defined by the difference of the lower subband signal. Because wavelet transform has the ability of reducing the noise, the derived features perform better. Our experiments show that higher rate of correct classification to SPOT image is obtained. In the meantime it shows that the texture's self-similarity is more evident in lower subband than that in gray level.

关 键 词：多尺度 子带算法 HURST参数 小波变换 纹理分类 遥感数据 分类 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]