摘要
为更好地描述具有细微运动图像序列的运动细节,揭示潜在运动信息,提出一种细微运动放大算法.设计了基于欧拉放大的图像序列时空处理方法,获得运动放大图像序列;提出基于视觉显著性的运动区域检测方法,获取准确的细微运动区域;最后设计了基于区域的运动细节增强方案,提高了运动区域的清晰度,减轻了运动放大所引入的瑕疵.以上述方法为基础,最终可获取高质量的运动放大图像序列.与最新的图像运动放大算法对比的实验结果表明,提出的算法更为准确地表征图像中对象的运动细节,同时在客观量化评价指标上取得了更好的效果.
In order to better describe the motion details of image sequences as well as reveal underlying motion information, the subtle motion magnification algorithm for image sequences is proposed. Firstly, the temporal and space processing approach based on Eulerian magnification is designed for acquiring motion amplification image sequences. Secondly, the motion region detection approach based on visual saliency is proposed for capturing subtle motion regions accurately. Finally, the motion detail enhancement scheme based on regions is designed, and it could improve the definition of motion regions of image sequences and reduce the introducing flaws due to motion enlarging. On this basis, the proposed algorithm could obtain high-quality motion amplification image sequences. Compared to the latest motion magnification algorithm, experiments show that the proposed algorithm is more accurate on describing the motion details of image sequences and achieves higher objective indexes.
出处
《小型微型计算机系统》
CSCD
北大核心
2016年第5期1066-1070,共5页
Journal of Chinese Computer Systems
基金
辽宁省教育厅科学研究一般项目(L2015225)资助
辽宁工程技术大学博士启动基金项目(14-1045
14-1126)资助
辽宁工程技术大学生产技术问题调研基金项目(14-T-009)资助
关键词
细微运动
欧拉放大
运动放大
视觉显著性
细节增强
subtle motion
eulerian magnification
motion magnification
visual saliency
detail enhancement