摘要
文章针对复杂背景下彩色图像中人脸检测算法复杂度大、计算时间长的问题,提出了一种改进方法。对输入图像进行自适应光照补偿,按色彩变化建立YCbCr肤色模型,筛选潜在人脸区域;对该区域进行Gabor滤波获得图像特征向量,用主成分析法对其降维;利用经训练的神经网络对候选区域进行判别。仿真实验表明该方法检测算法复杂度明显降低,精度可以达到95%以上。
Aim. The introduction of the full paper reviews some papers in the open literature and then, keeping in mind their two major shortcomings, proposes the research mentioned in the.title, which we believe to be new and better and which is explained in sections 1 thorough 6. The core of our explanation consists of： ＂with our algorithm, we can solve the problem of face detection in color images or in complex background. First, we make the self-adaptive lighting compensation, build the YCb Cr face model using nonlinear color transformation, and confirm the possible face region. Then we use Gabor filter for the possible face region to obtain image feature vector which will reduce the number of dimensions used in PCA （ principal component analysis） method. Lastly, we distinguish the possible region with neural network. ＂ Simulation results, presented in Figs. 7 and 8, in section 6, and the comparison of these results, presented in Table 1 in section 5, show preliminarily that our algorithm can indeed reduce complexity and improve precision; the examination precision reaches as high as 95 %.
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
2011年第5期690-694,共5页
Journal of Northwestern Polytechnical University
基金
博士点基金(博导类)(20096102110027)资助
关键词
GABOR滤波
肤色模型
神经网络
人脸检测
algorithms, feature extraction, independent component analysis, models, neural networks, simulation, face detection, face model, Gabor filter, principal component analysis（PCA）
