Abstract: In video communication with a human face, users mainly concentrate on the area of the face part. According to MARR vision principle, this paper proposes a method of facial video preprocessing based on region of interest, under the premise of ensuring the subjective quality of video communication. This method can effectively improve the efficiency of video communication. Firstly, it uses face detection algorithm to detect human face video. And then it utilizes bilateral filtering algorithm to handle background area outside of the face part. Finally it uses H.264 encoding algorithm for encoding transmission of processed video images. Experimental results demonstrate that the proposed method can effectively improve video coding efficiency, which can be up to 28.571%.

Key words: region of interest, bilateral filtering, face detection, H.264 encoding, compression ratio

摘要： 在头像视频通信中，网络用户的关注度一般集中在人脸部分区域。根据MARR视觉原理，在保证视频通信主观质量的前提下，提出了一种基于感兴趣区域的头像视频前处理方法，能有效地提高视频的通信效率。该方法首先利用人脸检测算法对头像视频进行人脸检测，然后利用双边滤波算法对人脸之外的背景区域进行滤波操作，最后用H.264编码算法对处理后的视频图像进行编码传输。实验结果表明，所提出的方法能够有效地提高视频编码效率，最高可以达到28.571%。

关键词: 感兴趣区域, 双边滤波, 人脸检测, H.264编码, 压缩率