计算机应用 ›› 2015, Vol. 35 ›› Issue (2): 535-539.DOI: 10.11772/j.issn.1001-9081.2015.02.0535

• 虚拟现实与数字媒体 • 上一篇    下一篇

基于局部约束邻域嵌入的人脸画像照片合成

胡彦婷1, 王楠楠2, 陈建军1,3, 木拉提·哈米提1, 阿布都艾尼·库吐鲁克1   

  1. 1. 新疆医科大学 医学工程技术学院, 乌鲁木齐 830011;
    2. 西安电子科技大学 通信工程学院, 西安 710071;
    3. 西南大学 物理科学与技术学院, 重庆 400715
  • 收稿日期:2014-08-04 修回日期:2014-09-16 出版日期:2015-02-10 发布日期:2015-02-12
  • 通讯作者: 木拉提·哈米提
  • 作者简介:胡彦婷(1980-),女,湖北孝感人,讲师,硕士,主要研究方向:图像超分辨重建、计算机视觉; 王楠楠(1986-),男,山东潍坊人,博士,主要研究方向: 异质图像转换、计算机视觉、机器学习; 陈建军(1977-),男,河南开封人,讲师,博士研究生,主要研究方向:光电信号处理; 木拉提·哈米提(1957-),男(维族),新疆乌鲁木齐人,教授,主要研究方向:图像处理及分析; 阿布都艾尼·库吐鲁克(1977-),男(维族),新疆喀什人,副教授,博士,主要研究方向:信息检测与处理。
  • 基金资助:

    国家自然科学基金资助项目(81160182,61201125)。

Face sketch-photo synthesis based on locality-constrained neighbor embedding

HU Yanting1, WANG Nannan2, CHEN Jianjun1,3, MURAT Hamit1, ABDUGHRNI Kutluk1   

  1. 1. School of Medical Engineering Technology, Xinjiang Medical University, Urumqi Xinjiang 830011, China;
    2. School of Telecommunications Engineering, Xidian University, Xi'an Shaanxi 710071, China;
    3. School of Physical Science and Technology, Southwest University, Chongqing 400715, China
  • Received:2014-08-04 Revised:2014-09-16 Online:2015-02-10 Published:2015-02-12

摘要:

针对画像块和照片块在流形上的邻域关系并不能完全反映彼此内在数据结构的问题,提出一种基于局部约束邻域嵌入(LCNE)的画像-照片合成算法。首先,利用基于邻域嵌入(NE)的合成方法得到待合成照片或画像的初始估计;其次,根据待合成的照片块或画像块与训练集中的照片块或画像块的相似性来约束合成权值;然后,通过交替优化方法进行权值的确定和K-近邻的选择,并更新待合成目标块;最后,合并所有估计的照片块或画像块合成目标图像。与基于邻域嵌入的画像照片合成方法相比,所提方法合成图像的结构相似度提高0.0503,脸识别准确率提高14%。实验结果表明,该方法解决了基于NE方法导致的邻域之间兼容性不强的问题,能大大减少合成图像上的噪声及变形。

关键词: 人脸画像-照片合成, 局部约束, 邻域嵌入, 流形学习, 人脸识别

Abstract:

The neighboring relationship of sketch patches and photo patches on the manifold cannot always reflect their intrinsic data structure. To resolve this problem, a Locality-Constrained Neighbor Embedding (LCNE) based face sketch-photo synthesis algorithm was proposed. The Neighbor Embedding (NE) based synthesis method was first applied to estimate initial sketches or photos. Then, the weight coefficients were constrained according to the similarity between the estimated sketch patches or photo patches and the training sketch patches or training photo patches. Subsequently, alternative optimization was deployed to determine the weight coefficients, select K candidate image patches and update the target synthesis patch. Finally, the synthesized image was generated by merging all the estimated sketch patches or photo patches. In the contrast experiments, the proposed method outperformed the NE based synthesis method by 0.0503 in terms of Structural SIMilarity (SSIM) index and by 14% in terms of face recognition accuracy. The experimental results illustrate that the proposed method resolves the problem of weak compatibility among neighbor patches in the NE based method and greatly alleviates the noises and deformations in the synthetic image.

Key words: face sketch-photo synthesis, local constraint, Neighbor Embedding (NE), manifold learning, face recognition

中图分类号: