基于自适应边缘阈值及方向加权的空间错误隐藏算法  

Spatial error concealment algorithm based on adaptive edge thresholding and directional weight

作　　者：李玉峰[1] 李广泽[2] 龙科慧[2] 

机构地区：[1]沈阳航空航天大学电子信息工程学院,辽宁沈阳110136 [2]中国科学院长春光学精密机械与物理研究所,吉林长春130033

出　　处：《光学精密工程》2016年第3期626-634,共9页Optics and Precision Engineering

基　　金：国家自然科学基金资助项目(No.61171081);辽宁省自然科学基金资助项目(No.2013024008)

摘　　要：针对H.264/AVC压缩视频码流在无线信道传输过程中会由于数据丢包导致图像重构质量下降的问题,提出了一种基于自适应边缘阈值及方向加权的空间错误隐藏算法。该算法利用图像边缘检测Sobel梯度算子检测相邻宏块边缘;根据受损宏块的相邻宏块具体信息自适应设定梯度阈值,最后对受损宏块进行方向加权插值从而重构图像。实验表明,该算法简单实用,不仅保留了丢失块像素加权平均算法的优点,而且能够用于边缘信息强度不同的错误隐藏。在不同的实时传输协议（RTP）丢包概率情况下,该算法的峰值信噪比较传统自适应算法提高了0.2-0.4 db,较多方向插值算法提升了0.2-3.8 db,提高了图像恢复质量,而且具有较高的应用价值。In wireless transmission, H.264/AVC compressed video stream shows poorer image reconstruction quality due to the data losing, so this paper presents a spatial error concealment algorithm based on adaptive edge thresholds and directional weights. This algorithm uses Sobel gradient operator in image edge detection to detect the edge of adjacent macroblocks. Then it sets adaptively gradient threshold according to the specific information of adjacent macroblock in the damaged macroblock. Finally, it makes the direction weighted interpolation for the damaged macroblock implement the image reconstruction. Experiments show that the algorithm is simple and practical. It not only retains the advantages of the weighted average algorithm and conceals the errors of the images with different edge information intensities. The Peak Signal to Noise Ratio of the algorithm has improved 0.2-3.8 db as compared to that of the Multi-Directional Interpolation algorithm in different probability events of data losing for Real-time Transport Protocol（RTP）. This algorithm improves the quality of image restoration and has a higher application value.

关 键 词：H.264/AVC 错误隐藏 自适应边缘阈值 方向加权 Sobel梯度算子 

分 类 号：TN943[电子电信—信号与信息处理]