自适应权重联合多尺度LBPV^2纹理分类方法

A adaptive weight joint multi-scale LBPV^2 for texture classification

传统的局部二值模式仅局限于局部纹理信息的提取,忽略了全局纹理信息的表达,造成最终的纹理分类效果并不理想。为了解决以上问题,借鉴局部二值模式方差(LBPV)的优势,在此基础上提出了一种新的基于自适应权重联合多尺度LBPV^2的纹理图像分类方法。该方法将方差平方作为直方图累积权重取代原来的方差权重,并采用自适应权重联合多尺度方案来实现多尺度纹理信息提取,进一步提升了纹理图像描述子的分类性能。在国际公认的Outex纹理数据集上的仿真实验表明,提出的这种新的基于自适应权重联合多尺度LBPV^2的纹理图像分类方法能够实现纹理分类性能的显著改善。

Local binary pattern(LBP)has been widely used in texture classification,however,traditional local binary pattern is limited in extracting local texture information,and it loses sight of the representation of global texture information,which make the texture classification task not doing well.In order to solve this problem,by taking advantage of local binary pattern variance(LBPV),this paper proposes a new texture image classification method based on the adaptive weight joint multi-scale LBPV^2.In this method,it considers the square of variance as histogram cumulative weight instead of traditional variance weight,and it also uses the adaptive weight joint multi-scale scheme to extract the multi-scale texture information,thereby the texture classification performance is further improved.The simulation experiments conducted on popular Outex benchmark texture database indicate that the proposed adaptive weight joint multi-scale LBPV^2(AWJLBPV^2)can greatly improve the texture classification performance.