Effect of Parameter Optimization on Classical and Learning-based Image Matching Methods

被引:4
|
作者
Efe, Ufuk [1 ]
Ince, Kutalmis Gokalp [1 ]
Alatan, A. Aydin [1 ]
机构
[1] Middle East Tech Univ, Dept Elect & Elect Engn, Ctr Image Anal OGAM, Ankara, Turkey
关键词
D O I
10.1109/ICCVW54120.2021.00283
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Deep learning-based image matching methods are improved significantly during the recent years. Although these methods are reported to outperform the classical techniques, the performance of the classical methods is not examined in detail. In this study, we compare classical and learning-based methods by employing mutual nearest neighbor search with ratio test and optimizing the ratio test threshold to achieve the best performance on two different performance metrics. After a fair comparison, the experimental results on HPatches dataset reveal that the performance gap between classical and learning-based methods is not that significant. Throughout the experiments, we demonstrated that SuperGlue is the state-of-the-art technique for the image matching problem on HPatches dataset. However, if a single parameter, namely ratio test threshold, is carefully optimized, a well-known traditional method SIFT performs quite close to SuperGlue and even outperforms in terms of mean matching accuracy (MMA) under 1 and 2 pixel thresholds. Moreover, a recent approach, DFM, which only uses pre-trained VGG features as descriptors and ratio test, is shown to outperform most of the well-trained learning-based methods. Therefore, we conclude that the parameters of any classical method should be analyzed carefully before comparing against a learning-based technique.
引用
收藏
页码:2506 / 2513
页数:8
相关论文
共 50 条
  • [1] PERFORMANCE EVALUATION OF LEARNING-BASED METHODS FOR MULTISPECTRAL SATELLITE IMAGE MATCHING
    Jovhari, N.
    Farhadi, N.
    Sedaghat, A.
    Mohammadi, N.
    ISPRS GEOSPATIAL CONFERENCE 2022, JOINT 6TH SENSORS AND MODELS IN PHOTOGRAMMETRY AND REMOTE SENSING, SMPR/4TH GEOSPATIAL INFORMATION RESEARCH, GIRESEARCH CONFERENCES, VOL. 10-4, 2023, : 335 - 341
  • [2] An evaluation of conventional and deep learning-based image-matching methods on diverse datasets
    Ji, Shunping
    Zeng, Chang
    Zhang, Yongjun
    Duan, Yulin
    PHOTOGRAMMETRIC RECORD, 2023, 38 (182): : 137 - 159
  • [3] LEARNING-BASED TONE MAPPING OPERATOR FOR IMAGE MATCHING
    Rana, Aakanksha
    Valenzise, Giuseppe
    Dufaux, Frederic
    2017 24TH IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING (ICIP), 2017, : 2374 - 2378
  • [4] Learning-based Ballistic Breech Face Impression Image Matching
    Roth, Joseph
    Carriveau, Andrew
    Liu, Xiaoming
    Jain, Anil K.
    2015 IEEE 7TH INTERNATIONAL CONFERENCE ON BIOMETRICS THEORY, APPLICATIONS AND SYSTEMS (BTAS 2015), 2015,
  • [5] Learning-Based Tone Mapping Operator for Efficient Image Matching
    Rana, Aakanksha
    Valenzise, Giuseppe
    Dufaux, Frederic
    IEEE TRANSACTIONS ON MULTIMEDIA, 2019, 21 (01) : 256 - 268
  • [6] A review of image matching methods based on deep learning
    Liu, Haiqiao
    Liu, Meng
    Gong, Zichao
    Dong, Jing
    Hangkong Xuebao/Acta Aeronautica et Astronautica Sinica, 2024, 45 (03):
  • [7] A Universal Optimization Framework for Learning-based Image Codec
    Zhao, Jing
    Li, Bin
    Li, Jiahao
    Xiong, Ruiqin
    Lu, Yan
    ACM TRANSACTIONS ON MULTIMEDIA COMPUTING COMMUNICATIONS AND APPLICATIONS, 2024, 20 (01)
  • [8] Parameter optimization of machining processes using teaching–learning-based optimization algorithm
    P. J. Pawar
    R. Venkata Rao
    The International Journal of Advanced Manufacturing Technology, 2013, 67 : 995 - 1006
  • [9] Depth from a Light Field Image with Learning-Based Matching Costs
    Jeon, Hae-Gon
    Park, Jaesik
    Choe, Gyeongmin
    Park, Jinsun
    Bok, Yunsu
    Tai, Yu-Wing
    Kweon, In So
    IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2019, 41 (02) : 297 - 310
  • [10] Using outlier elimination to assess learning-based correspondence matching methods
    Ding, Xintao
    Luo, Yonglong
    Jie, Biao
    Li, Qingde
    Cheng, Yongqiang
    INFORMATION SCIENCES, 2024, 659