提出了一种基于加权运动估计、矢量分割和可变块层次化处理的运动补偿内插(Motion-compensated frame inter-polation, MCFI)算法。首先,提出一种加权运动估计改善了运动矢量(Motion vector, MV)的准确度,其次,通过矢量聚类分割将视频帧分割为运动区域和背景,然后对运动区域的运动矢量进行了可变块层次化处理。此过程中,采用可变块合并算法保证了运动物体的边缘结构信息不被损坏。同时考虑到部分可变块的多方向性,使用了自适应矢量中值滤波器和矢量平滑降低了运动块大小,能有效地消除传统方法中出现的方块效应和重影现象。实验结果表明该算法在内插图像的主观视觉效果和客观评估标准上都有所提高,而且对于运动较快及背景较复杂的视频序列同样具有较强的适应性。%In this paper, an algorithm of motion-compensated frame interpolation is proposed based on weighted motion estimation, motion vector (MV) segmentation and variable block processing (MCFI). First, we present the weighted motion estimation method to obtain more accurate motion vectors. Next, we partition a frame into moving areas and background by clustering the received motion vectors. Then, a variable block processing method is applied to the moving areas, which can protect the moving areas0 edge structures from being damaged. Meanwhile, considering multiple directions of the motion blocks, we adopt adaptive vector median filter and motion vector smoothing to refine the block size to effectively eliminate the blocking artifacts and ghost effects. Experiment shows that the algorithm not only improves the image quality both subjectively and objectively, but also has a better adaptability to video sequences with fast motion and complex background.
展开▼
