电子学报 ›› 2022, Vol. 50 ›› Issue (6): 1451-1456.DOI: 10.12263/DZXB.20200664

• 学术论文 • 上一篇    下一篇

基于局部投影的视差图像拼接平滑优化

刘杰1, 游品鸿1, 田明2, 刘金凤1   

  1. 1.哈尔滨理工大学测控技术与通信工程学院,黑龙江 哈尔滨 150080
    2.中国电信黑龙江省分公司,黑龙江 哈尔滨 150080
  • 收稿日期:2020-07-08 修回日期:2021-01-04 出版日期:2022-06-25
    • 作者简介:
    • 刘 杰 女,1980年出生,黑龙江齐齐哈尔人.副教授.2005年于哈尔滨工业大学获得硕士学位,2013年于哈尔滨理工大学获得博士学位.主要研究方向为光纤光栅传感及解调技术、数字图像处理及计算机视觉技术.E-mail: liujie@hrbust.edu.cn
      游品鸿 男,1995年出生,福建福州人.哈尔滨理工大学测控技术与通信工程学院硕士研究生.主要研究方向为计算机视觉以及图像处理.E-mail: 1091453221@qq.com
    • 基金资助:
    • 国家自然科学基金 (51607049); 黑龙江省自然科学基金 (LH2019E067)

Smooth Optimization of Parallax Images Mosaic Based on Local Projection

LIU Jie1, YOU Pin-hong1, TIAN Ming2, LIU Jin-feng1   

  1. 1.School of Measurement and Control Technology and Communication Engineering, Harbin University of Science and Technology, Harbin, Heilongjiang 150080, China
    2.Heilongjiang Branch of China Telecom, Harbin, Heilongjiang 150080, China
  • Received:2020-07-08 Revised:2021-01-04 Online:2022-06-25 Published:2022-06-25
    • Supported by:
    • National Natural Science Foundation of China (51607049); Natural Science Foundation of Heilongjiang Province, China (LH2019E067)

摘要:

针对视差图像拼接之后存在对齐不准和扭曲的问题,提出一种基于局部投影的视差图像拼接的平滑优化方法.将输入图像进行特征点检测,通过局部保留结构进行图像匹配并计算投影矩阵,利用重合边界投影模型与投影矩阵进行归一化距离判定,距离小于阈值则保留,大于阈值则舍弃并继续矩阵计算.在图像非重合部分,利用图像的变形函数得到最小弯曲能量函数,使非重合区域向全局变换上过渡.实验结果表明:所提算法提高了图像匹配精度,对图像的扭曲具有优化作用,增强了图像的整体性并获得了更好的视觉体验.

关键词: 视差图像拼接, 投影矩阵, 全局变换, 阈值

Abstract:

Aiming at the problems of misalignment and distortion after parallax image stitching, a smooth optimization method for parallax image stitching based on local projection is proposed. The input image is performed feature point detection, matched by locality preserving structure, and calculated the projection matrix. The normalized distance is determined by the projection model of the coincident area and the projection matrix. If it is less than the threshold, it will be retained, otherwise it will be discarded and continue matrix calculation. In the non-coincident part of the image, the minimum bending energy function is obtained by calculating the deformation function of the image, so that the non-coincident region transits to the global transformation. The experimental results show that the proposed algorithm can not only improve the matching accuracy of the image, but also optimize the distortion of the image, enhance the integrity of the image and obtain a better visual experience.

Key words: parallax images stitching, projection matrix, global transformation, threshold

中图分类号: