固定场景下快速电子稳像算法的实现  

Realization of a fast algorithm for electron image stabilization under stationary scenes

作　　者：孙宇超[1] 张军[1] 

机构地区：[1]天津职业技术师范大学电子工程学院,天津300222

出　　处：《天津职业技术师范大学学报》2016年第1期22-26,共5页Journal of Tianjin University of Technology and Education

基　　金：天津市科技特派员项目(14JCTPJC00537)

摘　　要：提出一种可应用在固定场景条件下的快速电子稳像算法。通过检测参考图像与当前图像的ORB特征,利用汉明距离匹配特征,将匹配好的特征点代入仿射变换模型,求得2幅图像间的全局运动矢量并代入仿射变换模型,对当前图像进行变换,从而实现稳像目的。为达到更好的稳像效果,在图像预处理阶段对图像进行高斯滤波,滤除图像的噪声;同时对在求得全局运动矢量使用的特征点对的选择上采用点线法进行判定,保证求得全局运动矢量的精度。实验结果表明,该算法能够实现在固定场景下的稳像,并具有较好的实时性。A fast algorithm for electron image stabilization is presented which can be applied to the stationary scene conditions. The algorithm first detects the ORB features in the reference image and the current image and then matches the features by hamming distance. The matched features points are substituted for transformation model, so that the global motion vector between the two images is figured out. Finally, the global vector is applied to the current image for a affine transformation, and the images are stabilized. In order to achieve a better image stabilization, the images are filtered by Gauss filter and the noise of the images are filtered. In addition, the accuracy of the global motion vector is obtained by using the point-line method in the selection of the feature points. Experiments show that the algorithm can achieve stable image in a stationary scene and had a good real-time performance.

关 键 词：电子稳像 ORB特征 仿射变换模型 全局运动估计 固定场景 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]