CAJ | 학술논문

该文对人脸识别过程中的预处理进行研究,用高斯平滑、直方图均衡化和Sobel算子对图像进行预处理,通过改进前后在Feret人脸库的实验结果比较其识别率,观察各预处理方法对LBP算子性能的影响。大量的实验表明,结合高斯平滑预处理、Sobel与LBP的算法识别率提高最大。
해문대인검식별과정중적예처리진행연구,용고사평활、직방도균형화화Sobel산자대도상진행예처리,통과개진전후재Feret인검고적실험결과비교기식별솔,관찰각예처리방법대LBP산자성능적영향。대량적실험표명,결합고사평활예처리、Sobel여LBP적산법식별솔제고최대。
This paper do research for the pretreatment of face recognition process,do the image preprocessing with gaussian smooth,histogram equalization and Sobel operator,through improve the experiment results of before and after in Feret face library to compare the recognition rate,observe the influence from every pretreatment method to LBP operator performance.A large number of experiments show that,the biggest recognition rate increase is combined with gaussian smooth pretreatment,Sobel and the algorithm of LBP.