【摘要】 为提高电子稳像技术中运动矢量估计的准确性,采用基于兴趣区的特征点匹配进行运动矢量估计。该算法选取图像的兴趣区(ROI:Region of Interest)进行Harris的特征点检测,将ROI内的特征点稀疏化,利用SAD准则在相邻帧的特征窗内寻找匹配点对和距离准则剔除误匹配点进行逐帧运动补偿。实验结果验证了该算法的有效性。更多还原

【Abstract】 The key technology of electronic image stabilization is the motion vector estimation. FPS( Feature points) matching based on the ROI( Region of Interest) area is used to estimate motion vectors. Firstly,ROI of the input image was selected,then the FPS were detected by Harris. Secondly,the rule of SAD inside the window was used to sparse the FPS and find matching point in the adjacent frame. Finally,distance criterion was applied to eliminate false match points. Results of frame by frame motion compensation show the effectiveness of the algorithm.更多还原