改进型LBP算子在异源图像匹配的应用

Modified LBP Operator in Application of Multi-Source Image Matching

针对现有异源图像处理方法的不足,提出一种适合于描述异源图像纹理分布特征的改进的LBP特征的匹配算法。通过改进的LBP算子分别得到基准图和实时图的全局纹理分布特征;在此基础之上提取出基准图和实时图分块加权后的联合直方图特征并作为最终特征;并对图像分块加权后得到的联合LBP特征进行异源图像的特征匹配,在基准图上对此特征进行相似搜索得到相似系数图,峰值点即为匹配点。实验结果表明:通过LBP算子提取的纹理特征能够很好地应用于异源图像匹配,且匹配效果稳定优异,具有很好的工程应用价值。

Aiming at the deficiency of existing image processing method of multi-source, presents a matching algorithmwhich is suitable for the description of the improved LBP feature of multi-source image texture description. The referenceimage and real-time image of global texture distribution characteristics are obtained by using the improved LBP operator;based on the joint histogram features extracted from the reference image and real-time image block weighted and as thefinal feature; feature matching combined with LBP features and block weighted image of multi-source image, whichfeatures in the reference image similarity search coefficient map, the peak point is the matching point. The experimentalresults show that： the texture features extracted by LBP operator can be well used in multi-source image matching, andexcellent matching effect is stable, and has good engineering application value.