摘要
在传统的LBP算法的基础上,提出了一种改进的自适应阈值算法用于人类面部识别.提取图像的每个子区域的LBP,根据子区域图像自身的情况设定阈值,利用该阈值提取纹理特征,同时融合信息熵对分解的特征层进行直方图加权,在FERET人脸数据库上进行的实验证明,本文提出的算法具有更高的鉴别能力和对噪声干扰的更强的鲁棒性,能够有效提高图像检索的准确率.
Based on the traditional LBP algorithm, an improved adaptive threshold algorithm for human face recognition is proposed. Extraction of each sub region image LBP, it set the threshold according to the sub re-gion of the image itself, using texture feature extraction with the threshold value, at the same time integrated fu-sion information entropy histogram weighted on the decomposition of the feature layer, and demonstrated on FERET face database for experiments. The paper proposed that algorithm has higher ability to identify the im-ages and stronger robustness for noise interference which can effectively improve the accuracy of image re-trieval.
出处
《韶关学院学报》
2015年第2期11-14,共4页
Journal of Shaoguan University
基金
阜阳职业技术学院2013年教科研项目(2013JKYXM11)
作者简介
玉（1982-），女，安徽涡阳人，阜阳职业技术学院人文社科系讲师；研究方向：模式识别．
