DCT域快速下采样运动向量滤波器

Fast Motion Vector Filtering for Video Downsampling in the DCT Domain

目前的视频压缩标准多数采用DCT变换编码和运动补偿技术,运动估计约占整个编码时间的60%、运动补偿约占10%。所以在视频转码中,运动向量的再使用技术是十分重要的,目前较好的方法是欧氏最小距离方法,它的主要缺点是估计精度不高,本文对此进行了改进,提出了D CT域快速下采样运动向量滤波器,其重建图像的峰值信噪声比Shanableh等人提出的方法平均高0.2dB。

Most current video compression standards employ DCT transform and motion compensation,which approximately consumes 10% of encoding time, while motion estimation consumes about 60%. In trancoding,reusing the motion vectors extracted an incoming video bitsteam is very important. Better approach to reusing motion vectors is presently Euclidean minimizing distance with drawback of lower exactness. To improve it, we propose fast motion vector filtering for video downsampling in the DCT Domain with average of 0.2dB peak signal to noise ratio of reconstructed image gain over method proposed by Shanableh et al.