Journal of IKEEE (전기전자학회논문지)

Institute of Korean Electrical and Electronics Engineers (한국전기전자학회)
권호 표지
DOI QR코드

DOI QR Code

Comparison of error rates of various stereo matching methods for mobile stereo vision systems

모바일 스테레오 비전 시스템을 위한 다양한 스테레오 정합 기법의 오차율 비교

  • Received : 2022.12.12
  • Accepted : 2022.12.20
  • Published : 2022.12.31
Download PDF
⟨ Previous Next ⟩

Abstract

In this paper, the matching error rates of modified area-based, energy-based algorithms, and learning-based structures were compared for stereo image matching. Census transform (CT) based on region and life propagation (BP) algorithm based on energy were selected, respectively.Existing algorithms have been improved and implemented in an embedded processor environment so that they can be used for stereo image matching in mobile systems. Even in the case of the learning base to be compared, a neural network structure that utilizes small-scale parameters was adopted. To compare the error rates of the three matching methods, Middlebury's Tsukuba was selected as a test image and subdivided into non-occlusion, discontinuous, and disparity error rates for accurate comparison. As a result of the experiment, the error rate of modified CT matching improved by about 11% when compared with the existing algorithm. BP matching was about 87% better than conventional CT in the error rate. Compared to the learning base using neural networks, BP matching was about 31% superior.

본 논문에서는 스테레오 영상정합을 위하여 개선된 영역기반, 에너지 기반 알고리즘, 학습기반 구조의 정합 오류율을 비교하였다. 영역기반으로 census transform(CT), 에너지 기반으로 belief propagation(BP) 알고리즘을 선정하였다. 기존 알고리즘을 개선하고 모바일 시스템에서 스테레오 영상정합에 활용가능 하도록 임베디드 프로세서 환경에서 구현하였다. 비교 대상이 되는 학습기반의 경우에 도 적은 규모의 파라메터를 활용하는 신경망 구조를 채택하였다. 세 가지 정합방법의 오류율 비교를 위해 테스트 이미지로 Middlebury 데이터 세트 가운데 Tsukuba를 선정하고 정합 성능의 정확한 비교를 위해 비폐색, 불연속, 시차 오류율 등으로 세분화하였다. 실험 결과 CT 매칭의 오차율은 기존 알고리즘과 수정된 알고리즘으로 비교하였을 때 약 11% 성능 개선되었다. BP 매칭은 오류율에서 기존 CT 에 비하여 약 87% 우수하였다. 신경망을 이용한 학습기반과 비교 하였을 때 BP 매칭이 약 31% 우수함을 보였다.

Keywords

Acknowledgement

This work was supported by Seokyeong University in 2021.

References

  1. Ramin Zabih, John Woodfill, "Non-parmetric local transforms for computing visual correspondence," Eccv'94. LNCS, vol.801, pp.151-158, 2005. DOI: 10.1007/BFb0028345
  2. Aditya Rachman Putra, Fadhil Mochammad, Hans Herdian, "FPGA implementation of template matching using binary Sum of Absolute Difference," International Conference on System Engineering and Technology, 2016. DOI: 10.1109/ICSEngT.2016.7849615
  3. M. B. Hisham, Shahrul Niam Yaakob, R. A. A Raof, A. B A. Naren, N. M. Wafi, "Template Matching using Sum of Squared Difference and Normalized Cross Correlation," IEEE Student Conference on Research and Development, 2015. DOI: 10.1109/SCORED.2015.7449303
  4. J. N. Sarvaiya, Suprava Ptnaik, Salman Bombaywala, "Image Registration by Template Matching Using Normalized Cross-Correlation," International Conference on Advances in Computing, Control, and Telecommunication Technologies, 2009. DOI: 10.1109/ACT.2009.207s
  5. Chunsheng Guo, ShuoShuo Gao, Dazhuang Zhang, "Belief propagation algorithm for background estimation based on local maximum weight matching," International Congress on Image and Signal Processing, 2012. DOI: 10.1109/CISP.2012.6469678
  6. Ekaterina Mezenceva, Sergey Malakhov, "The study of the Semi-Global Block Matching Algorithm Implementing Parallel Calculation with GPU," International Conference on Problems of Infocommunications, Science and Technology, 2021. DOI: 10.1109/PICST54195.2021.9772166
  7. Zhao wei, Lian xiaofeng, Liu Zaiwen, "Panoramic image matching based on graph cuts algorithm," 2012 9th International Conference on Fuzzy Systems and Knowledge Discove. 2012. DOI: 10.1109/FSKD.2012.6234298
  8. J. Zbontar and Y. LeCun, "Stereo matching by training a convolutional neural network to compare image patches," Journal of Machine Learning Research, vol.17, no.1-32, pp.2, 2016. DOI: 10.5555/2946645.2946710
  9. H. Park and K. M. Lee, "Look wider to match image patches with convolutional neural networks," IEEE Signal Processing Letters, vol.24, no.12, pp.1788-1792, 2017. DOI: 10.1109/LSP.2016.2637355
  10. X. Ye, J. Li, H. Wnag, H. Huang, and X. Zhang, "Efficient stereo matching leveraging deep local and context information," IEEE Access, vol.5, pp.18745-18755, 2017. DOI: 10.1109/ACCESS.2017.2754318
  11. Hamid Laga, Laurent Valentin Jospin, Farid Boussaid, Mohammed Bennamoun, "A Survey on Deep Learning Techniques for Stereo-based Depth Estimation," IEEE Tran. on Pattern Analysis and Machine Intelligence, vol.44, no.4, pp.1738-1764, 2022. DOI: 10.1109/TPAMI.2020.3032602
  12. "Middlebury Stereo Datasets," http://vision.middlebury.edu/stereo/data/
  13. Jong il Park, Kyoung gyu Taek, Kwang yeob Lee, "Parallel Implementation of Improved Census Transform Stereo Matching algorithm with OpenMP," IEIE conference, vol.35, no.1,
  14. Jure Zbontar, Yann LeCun, "Stereo Matching by Training a Convolutional Neural Network to Compare Image Patches," The Journal of Machine Learning Research, vol.17, no.1, pp.2287-2318, 2016. DOI: 10.48550/arXiv.1510.05970
  15. "KITTI Stereo Data Set," http://https://www.cvlibs.net/datasets/kitti/
  16. I. Aleksander, "From WISARD to MAGNUS: A Family of Weightless Virtual Neural Machine," RAM-Based Neural Networks, World Scientific, pp.18-30, 1998. DOI: 10.1142/9789812816849_0002
  17. Martull Sarah, P. Martin, Fukui Kazuhiro, "Realistic CG Stereo Image Dataset With Ground Truth Disparity Maps," Computer Science, Environmental Science, 2012.
  18. Qian. Zhang, Shaomin Li, Y. Zhang, P. Wang, and JF. Huang, "New Stereo Matching Method based on Improved BP algorithm," International Journal on Smart Senging and Intelligent Systems, vol.8, no.1, 2015. DOI: 10.21307/ijssis-2017-767
  19. Lucas de Paula Veronese, Lauro Jose Lyrio Junior, Filipe All Mutz, Jorcy de Oliveira Neto, Vitor Barbirator Azevedo, Mariella Berger, Alberto Ferreira De Souza and Claudine Badue, "Stereo Matching with VG-RAM Weightless Neural Networks," Internaltional Conference on Intelligent Systems Design and Applications, pp.27-29 2012. DOI: 10.1109/ISDA.2012.6416556