基于改进的区域纹理谱描述符的图像检索

摘要

提出了区域纹理谱描述符的增强算法,新描述符在考虑局部邻域像素灰度差异符号变化的基础上,又融入了灰度差异的幅度变化以及局部邻域灰度均值与图像全局灰度均值的关系,从而更好地描述了局部区域纹理信息.为证明新描述符的分辨能力,文中采用不同图像库进行图像检索对比实验,实验结果证明新描述符的性能得到了较大程度的提高.%Based on local difference sign-magnitude transform,three completed modeling of the region local binary pattern ( LBP) operators ( completed center-symmetric LBP, completed direction LBP and completed variance and direction LBP) are proposed in the paper. The signs and the magnitudes of the image local differences and the gray-mean of the local region with the gray-mean of the whole image are coded together in the new descriptors. The traditional region LBP is part of the enhanced operator. The experimental results prove that the new operators show significant better performance than the traditional region LBPs for texture image retrieval.