人体头部图像移植算法研究

Algorithm research of Head Image Transplantation

本文针对将特定人体头部图象移植到视频图像序列中逐帧手工处理的状况,提出并实现一种相对定点位移的图像处理方法。首先采用差分图像法对视频图像中每帧图像及需要移植的特定图像进行人脸检测、定位,然后依据人脸特征并结合人脸肤色,找出特定图像人脸相对定点的位移量,根据位移量把特定人体头部图像移植到视频图像序列中去。实验证明该方法移植效果比较好,实现简单,高效。

According to the situation of transplanting image of a particular body's head to a video image frame by frame by hand in the past, a novel image processing method based on relatively fixed displacement is presented and experimented. Firstly, human face of each frame image of video images and particular image is detected and located by using difference image, then based on face fea-ture and skin color, relatively fixed displacement for particular image are detected. Finally , transplant particular head image to video image sequence by the above displacement. The experimental results show that the proposed method is successful and efficient.