数字稳像中的快速和鲁棒运动估计研究

Abstract

提出了一种用于数字图像稳定的快速和鲁棒运动估计方法.在图像的运动估计中,首先采用一种预判局部宏块的算法:在计算运动矢量前对宏块的梯度信息进行分析以避免误匹配,通过减少参与计算的宏块数目提高处理速度;其次,提出快速估计宏块运动矢量的方法,用改进的序贯相似性算法(SSdA)进行块匹配,以提高运动矢量的计算速度;然后采用lMEdS估计法去除不精确的宏块运动向量,用最小二乘法求解出最优的模型参数.实验结果表明了该方法的精确性、快速性和鲁棒性.

A fast and robust motion estimation method is proposed for digital image stabilization.First,an approach for block judgment is presented.Before motion vectors calculation,gradient information is analyzed to judge the blocks,only useful reference blocks that are indispensable for accurate motion estimation are selected with its reliability and consistency on pose estimation,by which the number of blocks for subsequent calculation is reduced.Second,in the block matching,an improved SSDA is used to reduce computing cost.Finally,using the LMedS method,wrong motion vectors of these blocks are eliminated and only reliable ones are used for whole motion estimation.Experimental results show the accuracy,fastness and robustness of the proposed method.