不同尺度的局部二元模式(LBP)提取了红外人脸图中不同的微结构局部特征。为了挖掘不同尺度中局部特征的相关性，提出了一种基于多尺度LBP共生直方图的红外人脸识别方法。传统的多尺度LBP特征提取方法，丢失了对多尺度特征间相关性信息的提取。为了充分考虑微结构间的相关统计信息，提出了多尺度LBP共生直方图表示方法，以提取包含在红外人脸图像中的有用鉴别特征。多尺度LBP共生直方图特征表示方法不仅可以消除环境温度对红外人脸图像特征提取的影响，而且还可以增强对局部特征表示的鉴别性。实验结果表明：多尺度局部二元模式共生矩阵可以增强对红外人脸鉴别特征提取的有效性，提出的红外人脸方法的性能优于基于传统多尺度LBP和单尺度LBP方法，在相同环境情况下和在环境温度变化情况下可以达到99.2%和91.2%的识别率。

Different scales local binary patterns(LBP) extract different micro-structures, which contain important discriminative information for infrared face recognition. To capture the correlation between different scales, a new infrared face recognition method based on multi-scale LBP co-occurrence histogram was proposed in this paper. In traditional multi-scale LBP-based features, correlation in different micro-structures was ignored. To consider such correlation in infrared faces, co-occurrence histogram of multi-scale LBP codes was used to represent the infrared face. Multi-scale LBP co-occurrence histogram not only preserved great invariance to environmental temperature, but also greatly enhanceed the discriminative power of the descriptor as co-occurrence matrix of LBP code well captureed the correlation between different scale micro-structures around the same central point. The experimental results show the recognition rates of infrared face recognition method based on multi-scale LBP co-occurrence histogram can reaches 99.2% under same condition and 91.2% under variable ambient temperatures, outperform that of the classic methods based on LBP and multi-scale LBP histogram.