基于自适应滤波的视频监控人脸超分辨率恢复  

Facial super⁃resolution recovery in video surveillance based on adaptive filtering

作　　者：陈铈 代荡荡 徐焕 CHEN Shi;DAI Dangdang;XU Huan(Information&Communication Company of State Grid Hubei Electric Power Co.,Ltd.,Wuhan 430000,China)

机构地区：[1]国网湖北电力有限公司信息通信公司,湖北武汉430000

出　　处：《电子设计工程》2025年第6期127-130,135,共5页Electronic Design Engineering

基　　金：国网湖北电力有限公司科研项目(0561XX02212)。

摘　　要：为了解决视频监控中噪声对人脸超分辨率恢复的影响,提出了基于自适应滤波的视频监控人脸超分辨率恢复方法。在计算自适应滤波权值后采用自适应滤波处理低分辨率图像,获取超分辨率近似图像。根据人脸特征空间坐标变换示意图检测人脸图像边缘信息。根据边缘信息将重建解限制在一个闭凸集中,获得最优解。根据最优解和反向投影获取人脸属性集,得到人脸超分辨率图像块行数和列数后,加权求和得到恢复后的超分辨率图像。实验结果表明,该方法峰值信噪比最大数值为46 dB,恢复结果清晰无噪点,说明该方法的人脸超分辨率恢复效果较好。In order to address the impact of noise on facial super-resolution recovery in video surveillance,a method based on adaptive filtering for facial super-resolution recovery in video surveillance is proposed.After calculating the adaptive filtering weights,the low resolution image is processed using adaptive filtering to obtain an approximate super-resolution image.Detect edge information of facial images based on the schematic diagram of coordinate transformation in facial feature space.Restrict the reconstruction solution to a closed convex set based on edge information to obtain the optimal solution.Obtain the facial attribute set based on the optimal solution and reverse projection,obtain the number of rows and columns of the facial super-resolution image blocks,and then weighted sum to obtain the restored super-resolution image.The experimental results show that the maximum peak signal-to-noise ratio of this method is 46 dB,and the restoration results are clear and free of noise,indicating that the facial super-resolution restoration effect of this method is good.

关 键 词：自适应滤波 视频监控人脸 超分辨率 恢复 

分 类 号：TN911[电子电信—通信与信息系统]