Unsupervised Deep Image Stitching: Reconstructing Stitched Features to Images

被引:128
作者
Nie, Lang [1 ,2 ]
Lin, Chunyu [1 ,2 ]
Liao, Kang [1 ,2 ]
Liu, Shuaicheng [3 ]
Zhao, Yao [1 ,2 ]
机构
[1] Beijing Jiaotong Univ, Inst Informat Sci, Beijing 100044, Peoples R China
[2] Beijing Key Lab Adv Informat Sci & Network Techno, Beijing 100044, Peoples R China
[3] Univ Elect Sci & Technol China, Sch Informat & Commun Engn, Chengdu 611731, Peoples R China
基金
中国国家自然科学基金;
关键词
Image stitching; Image reconstruction; Feature extraction; Image resolution; Estimation; Strain; Solid modeling; Computer vision; deep image stitching; deep homogrpahy estimation; HOMOGRAPHY;
D O I
10.1109/TIP.2021.3092828
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Traditional feature-based image stitching technologies rely heavily on feature detection quality, often failing to stitch images with few features or low resolution. The learning-based image stitching solutions are rarely studied due to the lack of labeled data, making the supervised methods unreliable. To address the above limitations, we propose an unsupervised deep image stitching framework consisting of two stages: unsupervised coarse image alignment and unsupervised image reconstruction. In the first stage, we design an ablation-based loss to constrain an unsupervised homography network, which is more suitable for large-baseline scenes. Moreover, a transformer layer is introduced to warp the input images in the stitching-domain space. In the second stage, motivated by the insight that the misalignments in pixel-level can be eliminated to a certain extent in feature-level, we design an unsupervised image reconstruction network to eliminate the artifacts from features to pixels. Specifically, the reconstruction network can be implemented by a low-resolution deformation branch and a high-resolution refined branch, learning the deformation rules of image stitching and enhancing the resolution simultaneously. To establish an evaluation benchmark and train the learning framework, a comprehensive real-world image dataset for unsupervised deep image stitching is presented and released. Extensive experiments well demonstrate the superiority of our method over other state-of-the-art solutions. Even compared with the supervised solutions, our image stitching quality is still preferred by users.
引用
收藏
页码:6184 / 6197
页数:14
相关论文
共 52 条
  • [41] PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volume
    Sun, Deqing
    Yang, Xiaodong
    Liu, Ming-Yu
    Kautz, Jan
    [J]. 2018 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2018, : 8934 - 8943
  • [42] Unsupervised Learning of Optical Flow With CNN-Based Non-Local Filtering
    Tian, Long
    Tu, Zhigang
    Zhang, Dejun
    Liu, Jun
    Li, Baoxin
    Yuan, Junsong
    [J]. IEEE TRANSACTIONS ON IMAGE PROCESSING, 2020, 29 (29) : 8429 - 8442
  • [43] Unsupervised Deep Homography: A Fast and Robust Homography Estimation Model
    Ty Nguyen
    Chen, Steven W.
    Shivakumar, Shreyas S.
    Taylor, Camillo Jose
    Kumar, Vijay
    [J]. IEEE ROBOTICS AND AUTOMATION LETTERS, 2018, 3 (03): : 2346 - 2353
  • [44] Deep Feature Extraction for Panoramic Image Stitching
    Van-Dung Hoang
    Diem-Phuc Tran
    Nguyen Gia Nhu
    The-Anh Pham
    Van-Huy Pham
    [J]. INTELLIGENT INFORMATION AND DATABASE SYSTEMS (ACIIDS 2020), PT II, 2020, 12034 : 141 - 151
  • [45] Wang L, 2020, PROCEEDINGS OF 2020 IEEE 4TH INFORMATION TECHNOLOGY, NETWORKING, ELECTRONIC AND AUTOMATION CONTROL CONFERENCE (ITNEC 2020), P694, DOI [10.1109/itnec48623.2020.9084886, 10.1109/ITNEC48623.2020.9084886]
  • [46] As-Projective-As-Possible Image Stitching with Moving DLT
    Zaragoza, Julio
    Chin, Tat-Jun
    Brown, Michael S.
    Suter, David
    [J]. 2013 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2013, : 2339 - 2346
  • [47] Zhang F, 2015, PROC CVPR IEEE, P2002, DOI 10.1109/CVPR.2015.7298811
  • [48] Parallax-tolerant Image Stitching
    Zhang, Fan
    Liu, Feng
    [J]. 2014 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2014, : 3262 - 3269
  • [49] Content-Preserving Image Stitching With Piecewise Rectangular Boundary Constraints
    Zhang, Yun
    Lai, Yu-Kun
    Zhang, Fang-Lue
    [J]. IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS, 2021, 27 (07) : 3198 - 3212
  • [50] Progressive Hard-Mining Network for Monocular Depth Estimation
    Zhang, Zhenyu
    Xu, Chunyan
    Yang, Jian
    Gao, Junbin
    Cui, Zhen
    [J]. IEEE TRANSACTIONS ON IMAGE PROCESSING, 2018, 27 (08) : 3691 - 3702