基于B-SURF算法的图像运动估计  

Image Motion Estimation Based on B-SURF Algorithm

作　　者：曾强 阿都建华[1] 周钦 邓成梁 ZENG Qiang;ADU Jianhua;ZHOU Qin;DENG Chengliang(Chengdu University of Information Technology,Chengdu,Sichuan 610225,China)

机构地区：[1]成都信息工程大学软件工程学院,四川成都610225

出　　处：《内江师范学院学报》2019年第2期51-56,66,共7页Journal of Neijiang Normal University

基　　金：国家自然科学基金(G030604);四川省科技厅重点项目(2018ZA0100);成都信息工程大学中青年学术带头人基金(J201709)

摘　　要：序列图像的运动矢量估计在电子稳像系统及图像超分辨率重建过程中都非常重要,传统的运动矢量估计算法大多需要提取相邻序列图像特征,根据相邻序列图像的特征点进行运动矢量估计,耗时相对较多,且难以达到实时性的要求.故提出并实现了一种基于SURF特征提取算法和BRIEF特征描述算子的运动估计方法:B-SURF方法,使用SURF特征提取算法思想对图像进行特征提取,采用BRIEF特征描述算子对检测到的特征点进行特征描述.由于引入了二进制编码,极大的降低了特征描述和匹配的计算量,提高了运动估计的计算速度,使算法能满足实时性需求.实验表明,实现方法在运算速度上有明显提升.Motion vector estimation for sequence images plays an important role in electronic image stabilization system and in the image super-resolution reconstruction process.The traditional motion vector estimation algorithm often has to extract the image characteristics of the adjacent sequences to estimate the motion vector according to the feature point of the adjacent sequence images,which is relatively time-consuming and difficult to achieve the real-time requirement.Therefore a motion estimation algorithm based on SURF feature extraction algorithm and BRIEF feature description operator is put forth:the B-SURF method,which adopts SURF algorithm for image feature extraction,and then uses the BRIEF description operator for character description of feature points that are detected.Owing to the introduction of binary coding,it drastically minimizes the amount of calculation of the character description and matching,and thus enhances the computation speed of motion estimation and makes the algorithm meet the real-time requirement.Experiments show that the method proposed in this paper has a significant improvement in the speed of computational operation.

关 键 词：运动估计 特征提取 特征匹配 SURF BRIEF 电子稳像系统 

分 类 号：TN919[电子电信—通信与信息系统]