运动补偿插帧是目前主要的帧率上转换方法.为减小内插帧中的“块效应”,并降低运算量以满足实时高清视频应用,该文提出了一种基于3维递归搜索(3-D Recursive Search,3-D RS)的多级块匹配运动估计视频帧率上转换算法.该算法将3-D RS与双向运动估计相结合,首先对序列中相邻帧进行“由粗到精”的三级运动估计,再利用简化的中值滤波器平滑运动矢量场,最后通过线性插值补偿得到内插帧.实验结果表明,与现有的运动补偿插帧算法相比,该算法内插帧的主、客观质量都有所提高,且算法复杂度低,有很强的实用性.%Motion-Compensated Frame Interpolation (MCFI) is a primary method for frame rate up conversion. A multi-stage block-matching motion estimation algorithm based on 3-D Recursive Search (3-D RS) is proposed in this paper to decrease the "block artifact" in the interpolated frame and to reduce the computational complexity for real-time high-definition video applications. In the proposed algorithm, combining 3-D RS with the bi-directional motion estimation, three-stage motion estimation for successive frames of the video sequence is firstly implemented step by step, and then the motion vector field is smoothened by using a simplified median filter. After that, the interpolated frame is produced via linear interpolation compensation. Compared with the existing algorithms of MCFI, the experimental results show that both objective and visual quality of the interpolated frame produced by the proposed algorithm is improved. It is feasible as well since its low complexity.
展开▼
