【摘要】 提出了一种自适应的运动补偿方案 ,首先利用预测三步搜索算法进行运动估计 ,由于此运动估计算法能够很好地利用运动向量的时空相关性 ,从而能得到更加平滑的运动向量场。其次 ,为了能够抑制运动向量场中个别奇异向量以及块匹配算法所固有的块效应问题 ,对向量场进行中值滤波 ,并采用了一种特殊的向量分配算法。方案的最后一个环节是运动补偿插值 ,它综合考虑插值帧的局部图像质量以及全局图像质量 ,利用一种自适应性较强的中值滤波操作来达到一定的综合效果。实验结果证明 ,此算法与运动补偿时间插值 (MCTI)算法相比 ,插值得到的图像更加平滑 ,而且 ,两者的信噪比指标对比也充分说明了本算法的优越性。更多还原

【Abstract】 A robust frame rate up-conversion scheme is proposed. Firstly a predictive three-step search method for performing motion estimation, the temporal and space correlation of motion vectors can be used and therefore a relatively smooth motion vector field is obtained. Then considering restraining the effects of some vectors and block artifacts of block matching algorithm, a median filter and particular vector assignment algorithm for motion vector field is proposed. The last part of the scheme is robust motion compensated interpolation, which can use a robust median filter to balance the local image quality and global image quality of the interpolated frames. Experimental results confirm that the interpolated frames from the algorithm are smoother than those from motion compensated temporal interpolation (MCTI), and the advantage of the algorithm can also be demonstrated.更多还原