基于帧间相关性的HDR视频压缩算法

Compression algorithm for HDR video based on interframe correlation

现有的高动态范围(HDR)的色调映射技术基本都是针对静态图像的,与视频相关的算法较少且效率不高。针对现有的HDR视频压缩算法,利用视频的帧间相关性,研究讨论了一种新的高效的压缩算法。单独设置参考帧作基于双边滤波的色调映射处理,将每帧图像分块与参考帧进行块匹配的运动估计,获取参考帧中的匹配块的低动态信息,避免了对视频序列中每帧图像都作色调映射处理。通过比较结构相似性和处理速度2方面对算法进行检验,实验结果表明:该算法在获取较好视觉效果的同时,大大加快了处理速度。

Existing tone mapping operators(TMOs)of high dynamic range are basically for static images and the related algorithms are limited and less efficient.Aiming at the existing HDR video compression algorithms,the interframe correlation of video is used and a new efficient compression algorithm is discussed.The reference frame is set for tone mapping processing based on bilateral filtering.The rest of frames are divided into some blocks and make a motion estimation matching with reference frame in order to get low dynamic information of matching blocks in the reference frame.It can avoid to process tone mapping on each frame of the video sequence.The algorithm is tested by comparing structural similarity and processing speed,experimental results show the algorithm greatly speeds up the processing while obtaining better visual effects.