运动视频序列影像的自动配准

影像配准是运动视频处理应用中的关键技术,从计算机视觉角度出发提出了一种针对运动视频处理的自动影像配准方法。其主要思路可以归结为基于特征匹配的配准过程,具体分为3个方面:利用Harris算子检测角点特征;以互相关函数为测度对角点特征进行初步匹配,特别使用RANSAC拟合基础矩阵F的方法剔除匹配中的错误对应;利用得到的结果重新拟合配准模型进行重采样变换。最后进一步分析了此方法应用于具体运动视频处理时需要考虑的一些问题。

Automatic Image Registration for Motion Video Sequence

Image registration is a key technique for some video processing applications.Based on computer vision theory,this paper presents an auto image registration method for motion video processing.The basic idea of it is a feature-based approach.The corner features were detected by Harris detector.And primary matching was processed using cross-correlation measurement.Then the outliers(false correspondence) were removed through the course of RANSAC fitting fundamental matrix,and re-fit the registration model using the results come from the method.At last,we make a further analysis of some questions that need to be considered when they used in motion video processing application.