系统仿真学报 ›› 2016, Vol. 28 ›› Issue (10): 2423-2430.

• 仿真建模理论与方法 • 上一篇    下一篇

基于ARM的视频流实时稳像增强算法

陈晓权, 张磊, 黄华   

  1. 北京理工大学计算机学院,北京 100081
  • 收稿日期:2016-05-28 修回日期:2016-07-12 出版日期:2016-10-08 发布日期:2020-08-13
  • 作者简介:陈晓权(1992-),男,浙江东阳,硕士生,研究方向为图像与视频处理;张磊(1981-),男,山东宁阳,副教授,研究方向为计算机图形学、图像与视频处理;黄华(1975-),男,湖北当阳,教授,研究方向为模式识别、图像与视频处理。
  • 基金资助:
    国家自然科学基金面上项目(61472035)

ARM-Based Real-Time Video Stabilization Enhancement Algorithm

Chen Xiaoquan, Zhang Lei, Huang Hua   

  1. School of Computer Science, Beijing Institute of Technology, Beijing 100081, China
  • Received:2016-05-28 Revised:2016-07-12 Online:2016-10-08 Published:2020-08-13

摘要: 针对ARM处理器的性能特点,提出一种在几何变换空间内进行视频抖动检测和运动平滑的高效视频稳像增强算法。首先通过平移和旋转分量的振幅变化检测视频流中出现的抖动;然后对平移向量和旋转角度进行线性插值,获得平滑路径的最优变换;最后分别对平移和旋转变换进行运动补偿,生成稳定的视频帧。由于采用刚性变换空间中的线性插值,该算法的计算速度要远高于以往算法,而且可以有效控制补偿后视频帧的裁剪率并保证不会出现画面扭曲现象。实验结果表明,该算法能够在基于ARM处理器的平台上实现视频流实时稳像处理,并输出令人满意的稳定视频。

关键词: 视频稳像, 实时, 刚性变换, 抖动检测

Abstract: Considering the performance characteristic of ARM processor, an ARM-based real-time video stabilization enhancement algorithm was proposed that identified shakiness of video and implements motion smoothing in the space of geometric transformations. The algorithm identified the shakiness of video with the amplitudes’ change of translation and rotation. Then the optimal path could be obtained by linear interpolation on translational vector and rotational angle. Stable frames were generated by implementing motion compensation of translation and rotation transformations respectively. Because the algorithm used linear interpolation in the space of rigid transformations, it was much faster than the previous algorithms. Moreover, it could control cropping ratio in motion compensation and guarantee no extra distortion drawn into the warped frames. Experiments show that the algorithm can stabilize video based on ARM platform in real time and obtain good video stabilization as well.

Key words: video stabilization, real-time, rigid transformation, shakiness detection

中图分类号: