目的 局部二进制模式(LBP)是一种理论简单、计算高效的非参数局部纹理特征描述子。由于其具有较高的特征鉴别力和较低的计算复杂度,因此近期获得了越来越多的关注,在图像分析、计算机视觉和模式识别领域得到了广泛的应用,尤其是在纹理分类和人脸识别两个经典的模式识别问题中,LBP方法得到充分的研究和发展。鉴于LBP的理论意义和实用价值,为了使国内外同行对LBP方法有一个较为全面的了解,对其进行系统总结。方法 在广泛文献调研的基础上,主要以纹理分类和人脸识别为应用背景,系统综述了LBP及现有各种LBP各种改进方法,从每种方法的研究动机、解决思路和方法特点及性能等方面进行总结。结果 首先,回顾了LBP方法的发展历程,综述了LBP及其众多改进方法的基本原理,系统梳理和评述了各种LBP方法的优势与不足,并在统一框架下对各种LBP方法进行分类总结;然后,综述了LBP及其各种改进方法在纹理分类和人脸识别中的应用研究,并总结了一些方法在基准数据库上达到的最高分类正确率;最后,凝练出LBP方法进一步的发展方向。结论 LBP方法的研究仍然是计算机视觉和模式识别领域倍受青睐的热点研究领域,仍然有更多低存储、快速的二值特征描述子被提出,LBP方法的应用领域仍在继续拓展。

Objective Local Binary Pattern (LBP) is a theoretically simple yet highly efficient texture descriptor. LBP has recently attracted increasing attention and has been successfully applied in image analysis, computer vision, and pattern recognition because of its discriminative power and computational simplicity. LBP has been developed for the traditional pattern recognition problems of texture classification and face recognition. Considering the theoretical and practical values of LBP, this study comprehensively reviews the suitability of various LBP variants in texture classification and face recognition. Method The fundamentals of the traditional LBP and various LBP variants are reviewed, and the advantages and disadvantages of various LBP variants are discussed by dividing them into categories under a novel systematic framework. Implicit directions for future LBP studies are also presented. Result First, the fundamentals of the traditional LBP method and various LBP variants are reviewed in detail and the pros and cons of various LBP variants are discussed by dividing them into different categories under a novel systematic framework. Second, as two typical and most successful applications of the LBP approach, LBP-based texture classification and LBP-based face recognition are reviewed. Finally, the implicit directions for future LBP research are presented. Conclusion LBP method continues to be a hot research topic in the field of computer vision and pattern recognition, which is evidenced by the fact that new low storage and fast local binary descriptors and new applications of LBP are still emerging.