一种改进的抗噪和旋转不变纹理图像分类算法  

Improved Antinoise and Rotation Invariant Texture Classification Algorithm

作　　者：尚燕[1] 安涛[1] 李春明[1] 

机构地区：[1]河北科技大学信息科学与工程学院电子系,河北石家庄050000

出　　处：《电视技术》2013年第9期38-41,共4页Video Engineering

基　　金：2011年度河北省科技厅科技计划项目(11213591);2011年度河北科技大学科学研究基金课题项目(XL201162)

摘　　要：直接在空域计算图像的局部二进制模式(LBP)直方图不能准确描述基元较大的纹理,且不具有抗噪性这一重要的特性。针对这一局限,提出一种将空域和频域结合起来的改进算法。首先在频域对图像进行Curvelet变换,选取能量较大的子带进行图像重构并抽样,最后计算抽样图像的旋转不变LBP算子直方图作为特征向量。这样得到的特征更加紧凑,能更有效地描述图像的细节,同时弥补了传统LBP算法的不足,实现了用相同尺寸的LBP算子描述大基元的纹理,同时具有旋转不变和抗噪性能。仿真结果表明,算法得到的分类率比其他算法有显著提高,并且具有较强的抗噪性。Computing the Local Binary Patterns（LBP） histogram of texture image in space directly can＇t character the texture whose texton is large and doesn＇ t have the performance of antinoise. Be aimed at these limitations, a improved algorithm which is in the combination of space and frequency domain is proposed. The texture image is transformed by Curvelet first, then compute the LBP histogram of the resampled image that is reconstructed using dominant directional subbands of each scale. The sparse rotation invariant feature vectors can characterize the detail of the texture and have the multi- resolution and anti-noise properties. At the same time , it remedied the limitation of LBP algorithm directly used in space and made the same sized LBP operators character the original texture in lager region. The proposed method is compared with other texture classification algorithm, the experiment results show that it can improve classification rate effectively and have stronger anti-noise properties.

关 键 词：多分辨率特征 抗噪 纹理分类 CURVELET变换 局部二进制模式 支持向量机 

分 类 号：TN911.73[电子电信—通信与信息系统]