H.264中基于检测与追踪的人脸保护方案

目前视频安全得到了人们的广泛关注,许多视频加密算法提出并得到了应用。但这些视频加密算法基本上是针对整帧图像的加密,并未考虑对图像中某些感兴趣的区域进行加密,往往增加了通信成本。针对此问题,提出了一种有效安全的感兴趣人脸区域保护方案,方案包括肤色信息和基于方向特征Bayes分类器相结合的人脸精确定位,基于Kalman滤波器的人脸预测跟踪以及H.264的CABAC编码中人脸区域的加密。实验结果表明,该方案不仅具有抗攻击性,而且安全有效地保护了感兴趣人脸区域。

Face protection of H.264 based on detecting and tracking

With more and more attention on security in video,people have proposed many video encryption algorithms and many of them have been put into application.However,these algorithms encrypted the whole frame,they didn't consider regions of interest,this would increase the communication cost.This paper proposed a safe protection scheme of interesting facial region.This approach included rough face detection based on skin-color model,accurate facial region location based on Bayes classifier depending on direction character convolved by Sobel filter,prediction based on Kalman filter and encrypt face region in CABAC coding.Experimental results show that this approach has better quality of anti-attack and protect interesting facial region effectively.