基于局部边缘差异二值模式的人脸识别方法

Face Recognition Based on Local Edge Difference Binary Patterns

传统的局部二值模式(local binary pattern,LBP)作为一种有效的特征提取与编码方法广泛应用于图像处理领域,但该方法只提取了图像模版中心像素值与边缘像素值的差值信息,这些信息不能全面地表征图像,并且对图像小幅度的灰度变化敏感.针对该方法的缺点,提出了一种基于局部边缘差异二值模式(local edge difference binary pattern,LEDBP)的特征提取算法,并应用于人脸识别领域.与传统的局部二值模式不同,本方法通过计算图像模版中心像素值与边缘像素值的差值来表示局部区域,其首先计算中心像素与相邻像素的边缘差值,然后使用局部三值模式的上模式(upper local ternary pattern)对边缘像素值进行编码,建立直方图后得到特征向量对图像进行表示.实验结果表明,本文算法用于人脸识别较传统的LBP等算法在识别率上有较大提高.

Traditional local binary pattern（LBP）is widely used in image processing as an effective feature extraction andencoding method,but it can only extract the different information between the central pixel and its neighbors in thetemplate,cannot fully represent the image,and is sensitive to small gray scale changes.According to the shortcomings ofthis method,a novel local edge difference binary pattern（LEDBP）is proposed for image feature extraction and used in theface recognition.The LEDBP differs from the existing LBP in a manner that it extracts information based on the distributionof edges in an image.The differences between the central pixel and its 8 neighbors in the template can be calculated by usingLEDBP.After that,the edges can be coded by using the upper local ternary pattern（LTPU）,and then the feature vectors canbe obtained by establishing histograms which represent the images.Experiments in face recognition show that the proposedmethod outperforms the traditional LBP on recognition accuracy.