基于SSIM的无线视频码率变化聚类识别算法

摘要

To dynamically recognize the bitrate variation in wireless video transmission process,according to the distribution characteristics of video frame quality,a SSIM based bitrate variation clustering recognition algorithm for wireless video named SBVCRA was proposed.Structural similarity (SSIM) between each pair of original and terminal decoding frames was used as the feature to quantify the degradation of video frame quality,and the similarity measure between two consecutive SSIM values was computed.Based on binary tree,the frame interval was recursively split using the frames with smallest value of similarity measure in each splitting process.Different video bitrates were clustered using SBVCRA,and the bitrate variation was recognized.The results show that the proposed algorithm can recognize the change of video bitrate by analyzing video frame quality and it is well consistent with the real bitrate changes in wireless video transmission.%为动态识别无线视频传输过程中视频码率变化情况,结合结构相似度(structural similarity,SSIM)的图像失真评价测度,根据终端视频帧质量数值分布特点,提出一种基于结构相似度的无线视频码率变化聚类识别算法SBVCRA(SSIM based bitrate variation clustering recognition algorithm for wireless video).使用结构相似度(SSIM)计算视频帧质量,计算相邻帧质量数据之间的相似程度;基于二叉树分裂方法,以相邻帧质量相似度值最小的帧序号为分裂点对帧区间进行递归分裂;依据分裂结果完成无线视频码率聚类,实现码率变化情况的识别.实验结果表明,所提算法可根据视频帧质量数值分布特性对相同码率自动聚类,码率变化情况的识别结果准确.