基于交互式感兴趣区域的教学视频编码

Interactive region-of-interest based lecture video encoding

为了在现有网络带宽和网络客户端处理能力的条件下提高网络多媒体教学系统中的视频质量,提出一种基于交互式感兴趣区域的教学视频编码算法。首先利用经典的GrabCut算法在低分辨率视频中交互式检测感兴趣目标,基于区域生长算法定位各目标所在区域并映射到高分辨率视频中;然后提取出感兴趣区域,缩放到显示分辨率,对各感兴趣区域编码生成H.264标准兼容的码流。实验结果表明,在相同视频质量下,该算法可以有效减小传输码率、降低客户端的计算复杂度,相比现存方法,传输码率最大节约94.3%,计算复杂度降低52.6%。

To improve the quality of lecture video in a network muhimedia teaching system under the condition of limited net- work bandwidth and clients＇ processing capacity, this paper proposed an encoding algorithm based on interactive region-of-in- terest （ROI）. Firstly, the algorithm adopted GrabCut algorithm to interactively detect the objects of interest from the low reso- lution video, and then used region growing algorithm to locate and map the objects to the high-spatial-resolution video. Then the algorithm extracted the ROIs and scaled them to display resolution and finally encoded them into stream with H. 264 standard. Experiment results indicate that this algorithm reduces transmission bitrate and computational complexity efficiently. Compared with existing methods, this algorithm saves at most 94.3 % of bitrates and reduces 52.6% of computational complexity for the same video quality.