基于特征提取的视频人脸图像受损区域修复研究  

Research on restoration of damaged area in video face image based on feature extraction

作　　者：马勇 MA Yong(Qinghai Police Vocational College,Xining 810000,China)

机构地区：[1]青海警官职业学院,青海西宁810000

出　　处：《电子设计工程》2023年第14期118-121,126,共5页Electronic Design Engineering

摘　　要：针对受损人脸图像在修复过程中易受到噪声影响的问题,采用特征提取方法,设计视频人脸图像受损区域修复方法。结合受损的图像纹理信息,判别图像受损区域。判断特征点到中心点的欧氏距离。提取图像边缘特征,设置分割线,将待修复的图像区域分成无边缘块和有边缘块,实现敏感受损区域特征分段修复。通过实验得出,该方法下图像的信噪比最高为42.354 dB,对左眼、右眼、鼻子、嘴巴特征点的修复效果均较好,能够得到一个最佳的修复结果。Aiming at the problem that the damaged face image is easy to be affected by noise in the repair process,a method of repairing the damaged area of video face image is designed by using feature extraction method.Combined with the damaged image texture information,the damaged area of the image is identified.Judge the Euclidean distance from the feature point to the center point.Extract the image edge features,set the segmentation line,and divide the image area to be repaired into non edge blocks and edge blocks,so as to realize the segmented repair of the features of sensitive damaged areas.The experimental results show that the maximum signal⁃to⁃noise ratio of the image under this method is 42.354 dB,and the repair effect of the feature points of the left eye,right eye,nose and mouth is good,which can get the best repair result.

关 键 词：特征提取 视频人脸图像 受损区域 修复 

分 类 号：TN108.1[电子电信—物理电子学]