QQ客服 官方微博 反馈留言

	基于噪点检测与邻域权值内插的彩色人脸图像去噪
	谢旭东 ; 袁兆君 ; 郭伟 ; 张毅 ; XIE Xudong ; YUAN Zhaojun ; GUO Wei ; ZHANG Yi
	2016-03-30 ; 2016-03-30
关键词	图像去噪 主成分分析 噪点检测 邻域权值内插 image denoising principle component analysis noisy pixel detection neighborhood weight interpolation TP391.41
其他题名	Color face image denoising based on noisy pixel detection and neighborhood weight interpolation
中文摘要	该文提出了一种基于彩色人脸图像训练库和Chebyshev不等式的有噪像素点检测算法,并利用像素间的相关性通过邻域权值内插算法,实现了彩色人脸图像的去噪。首先在L*a*b*颜色空间中进行主成分分析(principle component analysis,PCA)重建,然后定义了像素点向量距离差和像素点向量角度差,并给出有噪像素点判决准则,实现了有噪像素点与无噪像素点的区分,最后,利用像素点8邻域信息建立权值内插模型,结合PCA重建图像得到最终的去噪图像。在PIE标准人脸数据库中加入Gauss、椒盐、块噪声对该算法性能进行测试,并与其他几种去噪算法进行比较。实验结果表明:采用该算法去噪的彩色人脸图像具有更加清晰的边缘,保留了更多的有用信息,达到更高的峰值信噪比,且对多种噪声具有较好的去噪性能。; This paper presents an approach for color face image denoising based on noisy pixel point detection derived from face image databases using the Chebyshev Inequality and neighborhood weight interpolation.The first step is principle component analysis(PCA)in the L*a*b* color space for the face reconstruction.Then,apixel vector distance variance and a pixel vector angle variance are defined and a noisy pixel detection strategy is implemented.Finally,an eight-neighborhood weight interpolation model is used to get a PCA recovered image for the final denoised color face image.The algorithm performance is verified with Gaussian noise,pepper-and-salt noise and block noise added into the PIE standard color face database.The method is compared with several state-of-the-art image denoising methods.Tests based on the PIE database show that this approach is effective with different kinds of noise and achieves better performance with clearer edges,more useful information,better visual effects and higher PSNR.
语种	中文 ; 中文
内容类型	期刊论文
源URL	[http://ir.lib.tsinghua.edu.cn/ir/item.do?handle=123456789/147141]
专题	清华大学
推荐引用方式 GB/T 7714	谢旭东,袁兆君,郭伟,等. 基于噪点检测与邻域权值内插的彩色人脸图像去噪[J],2016, 2016.
APA	谢旭东.,袁兆君.,郭伟.,张毅.,XIE Xudong.,...&ZHANG Yi.(2016).基于噪点检测与邻域权值内插的彩色人脸图像去噪..
MLA	谢旭东,et al."基于噪点检测与邻域权值内插的彩色人脸图像去噪".(2016).

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。