基于非对称的三步搜索算法的运动估计
|
摘要
针对运动估计过程中传统的三步搜索算法的时间复杂度高的问题,提出了基于非对称的三步搜索算法的运动估计研究.首先将每帧图像按16×16的大小划分宏块作为压缩基本单元;然后,以当前宏块的中心点为中心,用非对称三步搜索算法搜索最佳匹配块的中心点,得到运动矢量;最后,对运动矢量进行编码作为压缩存储数据.实验表明:此算法与传统的全搜索、三步搜索算法相比,图像恢复后的信噪比相当;搜索时间比全搜索平均节约了67%,比三步搜索算法平均节约了40%;搜索点数比全搜索平均减少了89%,比三步搜索算法平均减少了6.5%.
Aiming at the high time complexity of the traditional three-step search algorithm in the motion estimation process, a motion estimation research based on the asymmetric three-step search algorithm is proposed. First, the image is divided into 16×16 macroblocks as the basic unit of compression. Second, center point of the current macroblock is taken as the center, and the center point of the best matching block is searched using the asymmetric three-step search algorithm to obtain the motion vector. Finally, the Motion vectors are encoded as compressed storage data. Experiments show that compared with the traditional full-search and three-step search algorithms, the proposed algorithm has a similar SNR after image recovery. The search time is saved by 67% compared with the full search, and it is saved by an average of 40% compared with the three-step search algorithm. The number of search points has been reduced by an average of 89% compared with the full search, and has been reduced by an average of 6.5% compared with the three-step search algorithm.
Aiming at the high time complexity of the traditional three-step search algorithm in the motion estimation process, a motion estimation research based on the asymmetric three-step search algorithm is proposed. First, the image is divided into 16×16 macroblocks as the basic unit of compression. Second, center point of the current macroblock is taken as the center, and the center point of the best matching block is searched using the asymmetric three-step search algorithm to obtain the motion vector. Finally, the Motion vectors are encoded as compressed storage data. Experiments show that compared with the traditional full-search and three-step search algorithms, the proposed algorithm has a similar SNR after image recovery. The search time is saved by 67% compared with the full search, and it is saved by an average of 40% compared with the three-step search algorithm. The number of search points has been reduced by an average of 89% compared with the full search, and has been reduced by an average of 6.5% compared with the three-step search algorithm.
引文
[ 1 ] 魏文振.基于H.264的帧间预测运动估计技术研究[D].合肥:合肥工业大学,2015.WEI W Z.Research on inter-frame prediction motion estimation technology based on H.264[D].Heifei:Hefei University of technology,2015.
[ 2 ] 孙晓亮,田力威,刘洋.基于改进的人工蜂群算法在医学图像上的多阈值图像分割法[J].沈阳大学学报(自然科学版),2017,29(6):479-484.SUN X L,TIAN L W,LIU Y.Multi-threshold image segmentation based on improved artificial bee colony algorithm on medical images[J].Journal of Shenyang University(natural science),2017,29(6):479-484.
[ 3 ] 赵春晖,李思聪.基于多特征融合的SVM高分辨率遥感影像建筑物提取方法[J].沈阳大学学报(自然科学版),2017,29(4):314-319.ZHAO C H,LI S C.SVM high-resolution remote sensing image building extraction method based on multi-feature fusion[J].Journal of Shenyang University(natural science),2017,29(4):314-319.
[ 4 ] 薛亮,于敏,张正炳.一种改进的运动估计算法:新三步搜索法[J].电视技术,2004(4):22-23.XUE L,YU M,ZHANG Z B.An improved motion estimation algorithm:new three-step search method[J].Television technology,2004(4):22-23.
[ 5 ] 吴通.基于H.264块匹配运动估计的研究[D].武汉:武汉理工大学,2008.WU T.Research on motion estimation based on H.264 block matching[D].Wuhan:Wuhan University of Technology,2008.
[ 6 ] 俞文静,张明军,王影.一种基于图像背景减除的菱形运动估计算法[J].计算机技术与发展,2017,27(4):42-45.YU W J,ZHANG M J,WANG Y.A diamond motion estimation algorithm based on image background subtraction[J].Computer technology and development,2017,27(4):42-45.
[ 7 ] 闫博冉,何卫锋,毛志刚.基于HEVC六边形运动估计算法的VLSI设计[J].电子科技,2017,30(9):130-134.YAN B R,HE W F,MAO Z G.A VLSI architecture of hexagonal search motion estimation for HEVC[J].Electronic technology,2017,30(9):130-134.
[ 8 ] 杨清永,王新,周俊辉.块运动估计中具有可扩展性的高精度三步搜索算法[J].西安电子科技大学学报,2002(5):610-613.YANG Q Y,WANG X,ZHOU J H.A highly accurate three-step search algorithm with extensibility for block motion estimation[J].Journal of Xidian University,2002(5):610-613.
[ 9 ] 文俊,王朋,刘重庆.一种预测三步搜索算法[J].上海交通大学学报,2003(6):862-865.WEN J,WANG P,LIU C Q.A new predictive three-step search algorithm[J].Journal of Shanghai Jiaotong University,2003(6):862-865.
[10] 张长帅,刘岩,陈祥滨,等.基于模糊逻辑的三步搜索算法[J].自动化与仪器仪表,2016(10):41-44.ZHANG C S,LIU Y,CHANG X B,et al.Three step search algorithm based on fuzzy logic[J].Automation and instrumentation,2016(10):41-44.
[11] 卢文涛,樊滨温.基于H.264运动估计的空间预测算法[J].计算机工程,2010,36(7):236-238.LU W T,FAN B W.Spatial prediction algorithm based on motion estimation in H.264[J].Computer engineering,2010,36(7):236-238.
[12] 刘书平.H.264运动估计算法比较与分析[J].现代计算机(专业版),2017(9):41-46.LIU S P.Comparison and analysis of H.264 motion estimation algorithms[J].Modern computer (professional edition),2017(9):41-46.
[ 2 ] 孙晓亮,田力威,刘洋.基于改进的人工蜂群算法在医学图像上的多阈值图像分割法[J].沈阳大学学报(自然科学版),2017,29(6):479-484.SUN X L,TIAN L W,LIU Y.Multi-threshold image segmentation based on improved artificial bee colony algorithm on medical images[J].Journal of Shenyang University(natural science),2017,29(6):479-484.
[ 3 ] 赵春晖,李思聪.基于多特征融合的SVM高分辨率遥感影像建筑物提取方法[J].沈阳大学学报(自然科学版),2017,29(4):314-319.ZHAO C H,LI S C.SVM high-resolution remote sensing image building extraction method based on multi-feature fusion[J].Journal of Shenyang University(natural science),2017,29(4):314-319.
[ 4 ] 薛亮,于敏,张正炳.一种改进的运动估计算法:新三步搜索法[J].电视技术,2004(4):22-23.XUE L,YU M,ZHANG Z B.An improved motion estimation algorithm:new three-step search method[J].Television technology,2004(4):22-23.
[ 5 ] 吴通.基于H.264块匹配运动估计的研究[D].武汉:武汉理工大学,2008.WU T.Research on motion estimation based on H.264 block matching[D].Wuhan:Wuhan University of Technology,2008.
[ 6 ] 俞文静,张明军,王影.一种基于图像背景减除的菱形运动估计算法[J].计算机技术与发展,2017,27(4):42-45.YU W J,ZHANG M J,WANG Y.A diamond motion estimation algorithm based on image background subtraction[J].Computer technology and development,2017,27(4):42-45.
[ 7 ] 闫博冉,何卫锋,毛志刚.基于HEVC六边形运动估计算法的VLSI设计[J].电子科技,2017,30(9):130-134.YAN B R,HE W F,MAO Z G.A VLSI architecture of hexagonal search motion estimation for HEVC[J].Electronic technology,2017,30(9):130-134.
[ 8 ] 杨清永,王新,周俊辉.块运动估计中具有可扩展性的高精度三步搜索算法[J].西安电子科技大学学报,2002(5):610-613.YANG Q Y,WANG X,ZHOU J H.A highly accurate three-step search algorithm with extensibility for block motion estimation[J].Journal of Xidian University,2002(5):610-613.
[ 9 ] 文俊,王朋,刘重庆.一种预测三步搜索算法[J].上海交通大学学报,2003(6):862-865.WEN J,WANG P,LIU C Q.A new predictive three-step search algorithm[J].Journal of Shanghai Jiaotong University,2003(6):862-865.
[10] 张长帅,刘岩,陈祥滨,等.基于模糊逻辑的三步搜索算法[J].自动化与仪器仪表,2016(10):41-44.ZHANG C S,LIU Y,CHANG X B,et al.Three step search algorithm based on fuzzy logic[J].Automation and instrumentation,2016(10):41-44.
[11] 卢文涛,樊滨温.基于H.264运动估计的空间预测算法[J].计算机工程,2010,36(7):236-238.LU W T,FAN B W.Spatial prediction algorithm based on motion estimation in H.264[J].Computer engineering,2010,36(7):236-238.
[12] 刘书平.H.264运动估计算法比较与分析[J].现代计算机(专业版),2017(9):41-46.LIU S P.Comparison and analysis of H.264 motion estimation algorithms[J].Modern computer (professional edition),2017(9):41-46.