李帅等：序列图像运动自适应VI-MT光流估计算法 ·1243· 方法，采用图像分解得到的纹理图部分计算光流，使计 ceedings of the IEEE Conference on Computer Vision and Pattern 算准确性显著提高.此外，针对FFVIMT光流估计算 Recognition.Boston,2015:1164 法权值选取问题，本文提出通过样本学习训练估计，有 [5]Li S,Xu Y L,Ma S P,et al.New method for SAR occluded tar- 效避免了选取函数拟合效果无法确定的不足.最后， gets recognition using DNN.J Xidian Unir,2015,42(3):154 (李帅，许悦雷，马时平，等.一种深度神经网路SAR遮挡目 利用由粗到精的金字塔多尺度分层优化框架，进一步 标识别方法.西安电子科技大学学报，2015,42(3)：154) 改善了估计效果.实验结果有力地表明了所提算法在 [6]Heeger D J.Optical flow using spatiotemporal filters.Int J Com- 运动估计上具有较好的效果.但是本研究中尽管Md put Vision,1988,1(4):279 dlebury标准库数据的随机性和普适性，实验中权值训 [7] Rust N C,Mante V,Simoncelli E P,et al.How MT cells analyze 练所用数据还有待扩展，而且生物细胞的数学模型还 the motion of visual patterns.Nat Neurosci,2006,9(11):1421 可以进一步深入分析和研究，如可以结合V2细胞的 [8]Nishimoto S,Gallant J L.A three-dimensional spatiotemporal re ceptive field model explains responses of area MT neurons to natu- 纹理提取特性取代STD技术等. ralistic movies.J Neurosci,2011,31(41):14551 [9]Solari F,Chessa M,Medathati N VK,et al.What can we expect 参考文献 from a V1-MT feedforward architecture for optical flow estima- tion?.Signal Processing:Image Commun,2015,39:342 [1]Fortun D,Bouthemy P,Kervrann C.Optical flow modeling and [10]Wedel A,Pock T.Zach C.et al.An improved algorithm for TV-L! computation:a survey.Comput Vision Image Understanding, optical flow /Statistical and Geometrical Approaches to Visual 2015,134:1 Motion Analysis.Berlin,2009:23 [2]Hom B K P,Schunck B G.Determining optical flow.Artif Intelli- [11]Ai J W,Liu K.A method of tracking a moving object in video gence,1981,17(1-3):185 sequences.J Univ Sci Technol Beijing,2006,28(2):195 [3]Butler D J,Wulff J,Stanley G B,et al.A naturalistic open (艾金慰，刘克.视频序列中运动目标跟踪新方法.北京科 souree movie for optical flow evaluation /12th European Confer- 技大学学报.2006,28(2)：195) ence on Computer Vision.Firenze,2012:611 [12]Baker S,Scharstein D,Lewis J P,et al.A database and evalua- [4]Revaud J.Weinzaepfel P,Harchaoui Z,et al.Epicflow:edge- tion methodology for optical flow.Int J Comput Vision,2011,92 preserving interpolation of correspondences for optical flowPro- (1):1李 帅等: 序列图像运动自适应 V1鄄鄄MT 光流估计算法 方法,采用图像分解得到的纹理图部分计算光流,使计 算准确性显著提高. 此外,针对 FFV1MT 光流估计算 法权值选取问题,本文提出通过样本学习训练估计,有 效避免了选取函数拟合效果无法确定的不足. 最后, 利用由粗到精的金字塔多尺度分层优化框架,进一步 改善了估计效果. 实验结果有力地表明了所提算法在 运动估计上具有较好的效果. 但是本研究中尽管 Mid鄄 dlebury 标准库数据的随机性和普适性,实验中权值训 练所用数据还有待扩展,而且生物细胞的数学模型还 可以进一步深入分析和研究,如可以结合 V2 细胞的 纹理提取特性取代 STD 技术等. 参 考 文 献 [1] Fortun D, Bouthemy P, Kervrann C. Optical flow modeling and computation: a survey. Comput Vision Image Understanding, 2015, 134: 1 [2] Horn B K P, Schunck B G. Determining optical flow. Artif Intelli鄄 gence, 1981, 17(1鄄3): 185 [3] Butler D J, Wulff J, Stanley G B, et al. A naturalistic open source movie for optical flow evaluation / / 12th European Confer鄄 ence on Computer Vision. Firenze, 2012: 611 [4] Revaud J, Weinzaepfel P, Harchaoui Z, et al. Epicflow: edge鄄 preserving interpolation of correspondences for optical flow / / Pro鄄 ceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Boston, 2015: 1164 [5] Li S, Xu Y L, Ma S P, et al. New method for SAR occluded tar鄄 gets recognition using DNN. J Xidian Univ, 2015, 42(3): 154 (李帅, 许悦雷, 马时平, 等. 一种深度神经网络 SAR 遮挡目 标识别方法. 西安电子科技大学学报, 2015, 42(3): 154) [6] Heeger D J. Optical flow using spatiotemporal filters. Int J Com鄄 put Vision, 1988, 1(4): 279 [7] Rust N C, Mante V, Simoncelli E P, et al. How MT cells analyze the motion of visual patterns. Nat Neurosci, 2006, 9(11): 1421 [8] Nishimoto S, Gallant J L. A three鄄dimensional spatiotemporal re鄄 ceptive field model explains responses of area MT neurons to natu鄄 ralistic movies. J Neurosci, 2011, 31(41): 14551 [9] Solari F, Chessa M, Medathati N V K, et al. What can we expect from a V1鄄鄄 MT feedforward architecture for optical flow estima鄄 tion?. Signal Processing: Image Commun, 2015, 39: 342 [10] Wedel A, Pock T, Zach C, et al. An improved algorithm for TV鄄鄄L 1 optical flow / / Statistical and Geometrical Approaches to Visual Motion Analysis. Berlin, 2009: 23 [11] Ai J W, Liu K. A method of tracking a moving object in video sequences. J Univ Sci Technol Beijing, 2006, 28(2): 195 (艾金慰, 刘克. 视频序列中运动目标跟踪新方法. 北京科 技大学学报, 2006, 28(2): 195) [12] Baker S, Scharstein D, Lewis J P, et al. A database and evalua鄄 tion methodology for optical flow. Int J Comput Vision, 2011, 92 (1): 1 ·1243·