一种快速的自动视频抠像算法研究  

ON A FAST AUTOMATED VIDEO MATTING ALGORITHM

作　　者：夏倩[1] 夏玉勤[2] 许勇[3] 

机构地区：[1]贵州民族大学传媒学院,贵州贵阳520025 [2]贵州大学计算机科学与工程学院,贵州贵阳520025 [3]华南理工大学计算机学院,广东广州510510

出　　处：《计算机应用与软件》2015年第10期201-204,共4页Computer Applications and Software

基　　金：贵州省科学技术基金项目（LKM[2012]11）.

摘　　要：传统的视频抠像算法，交互频繁，计算复杂度高。针对这些问题，利用从Kineet获得的深度图信息，彩色信息以及视频中运动信息，提出一种快速的自动视频抠像算法。算法首先采用改进的三帧间差分法检测到视频中的感兴趣区域（ROI），再根据深度图信息改进自动种子增长区域算法，估算出粗略的前景掩膜。接着，通过数学形态学和逻辑操作去除粗前景区域的杂点，自动生成大致的三分图。最后，采用改进了Shared Matting算法，得到精确的视频抠图。实验结果表明该算法不仅避免了人工交互，而且精度高、速度快。Traditional video matting algorithms ]lave frequent human interaction and high computation complexity. To solve these problems, we present a fast automated image matting algorithm using depth map information, colour information and the motion information in video retrieved from Kinect. First, the algorithm uses the improved three-frame difference method to detect the regions of interest （RO[） in video. Then by improving the automatic seed growing region algorithm according to depth map information, it estimates the rough foreground mask. After that, it removes the noises in rough foreground region through mathematical morphological and logical operations, and automatically generates an approximate trimap. Finally, it uses improved Shared Matting algorithm to obtain preeise video matting image. Experimental results show that the algorithm avoids the human interaction, and also has high precision and fast implementation.

关 键 词：感兴趣区域(ROI) 前景掩膜 三分图 视频抠图 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]