FOREX Press I. J. of Electrical & Electronics Research
Support Open Access
  • Home/
  • IJEER-140316

Research Article |

Depth Map Reconstruction in Underwater Robotic Vision using Gradient Spanning Trees and Boundary Segmentation

Author(s): Nurliyana Abd Mutalib 1, Rostam Affendi Hamzah 2*, Ken Prameswari Caesarella Aryaputri3, Ahmad Fauzan Kadmin4, Masrullizam Mat Ibrahim4, and Adi Irwan Herman 5

Published In : International Journal of Electrical and Electronics Research (IJEER)        Volume 14, Issue 3

Publisher : FOREX Publication

Published : 30 September 2026

e-ISSN : 2347-470X

Page(s) : 822-831

DOI: https://doi.org/10.37391/IJEER.140316

Abstract

Underwater robotic vision robot encounters͏ multiple challenges. Light is absorbed and scattered resulting in depth perception difficulties. The complexity of underwater scenes is worsened due to low texture, plain color, and irregular object boundaries. Hence, this article proposes a new algorithm to reconstruct disparity maps or depth maps from a stereo vision system used for underwater robotic vehicles. Fundamentally, there are four stages to develop the algorithm which are matching cost computation, cost aggregation, depth selection and depth refinement. To initiate the process a preliminary depth map is obtained using a bit-wise census-based matching cost computation. A high efficiency level is shown by this method even with differing lighting or colors on images. Cost aggregation employs a weighting approach from gradient Spanning Tree (ST) which applies gradient variations for enhancing the initial depth map. This method integrates gradient information into the spanning tree construction allowing the algorithm to preserve boundary discontinuities more effectively, reducing the risk of smoothing across object edges, unlike the common ST approaches that rely solely on edge weights. Next stage is depth selection which adopts a "winner-takes-all" (WTA) strategy. The depth value is determined by selecting minimum cost volume from normalization process. The final stage, boundary-based segmentation is proposed to refine final depth map. This method reconstructs ill-posed depth maps and increases the accuracy. Additionally, textures are added to object edges during this step resulting in precision of the final depth maps. Unlike prior ST-based and segmentation approaches, the proposed method uniquely integrates gradient-weighted spanning tree aggregation with boundary segmentation refinement, specifically proposed to overcome underwater vision challenges and complexities. Evaluations of the standard Middlebury system attains a non-occ error of 7.59%, all error of 10.90% and end-point-error at 0.5-1.0 demonstrating competitive performance compared with existing methods such as Esmea, PPEP-GF, Dense-CNN, ACMC and RDNet . The real images from the underwater Caddy benchmarking dataset further demonstrate that the proposed method produces clearer and more visually accurate results.

Keywords: Boundary Segmentation, Depth Map, Gradient, Spanning Tree, Underwater Vision.




Nurliyana Abd Mutalib, Engineering Technology Department, Universiti Teknikal Malaysia Melaka, 76100 Melaka, Malaysia

Rostam Affendi Hamzah ,Engineering Technology Department, Universiti Teknikal Malaysia Melaka, 76100 Melaka, Malaysia

Ken Prameswari Caesarella Aryaputri , Engineering Technology Department, Universiti Teknikal Malaysia Melaka, 76100 Melaka, Malaysia

Ahmad Fauzan Kadmin, Engineering Department, Universiti Teknikal Malaysia Melaka, 76100 Melaka, Malaysia

Masrullizam Mat Ibrahim, Engineering Department, Universiti Teknikal Malaysia Melaka, 76100 Melaka, Malaysia

Adi Irwan Herman, Texas Instruments Sdn Bhd, 74000 Melaka, Malaysia

    [1] G. Lu, P. Zhang, B. Song,” Stereo Matching Algorithm Combining Image Segmentation and Improved Belief Propagation”, Computer Fraud and Security, vol. 12, pp. 12–23, 2024, doi: 10.52710/CFS.273.
    [2] T. Li, Z. Huang, L. He, J. Wang, “Stereo Matching Algorithm Based on Improved Census Transform and Adaptive Window Optimization”, International Conference on Computer Science and Blockchain, 2024, pp. 219–224, doi: 10.1109/CCSB63463.2024.10735587.
    [3] M. Zahari, RA. Hamzah, “Stereo matching algorithm for autonomous vehicle navigation using integrated matching cost and non-local aggregation”, Bulletin of Electrical Engineering and Informatics, vol. 12, no 1, 2023, pp. 328–337, doi: 10.11591/EEI.V12I1.4122.
    [4] C. Lv, J. Li, S. Tang, ”Stereo Matching Algorithm Based on HSV Color Space and Improved Census”, Transform. Math Probl Eng, vol. 2021, no. 1, 2021, 1857327, doi: 10.1155/2021/1857327.
    [5] Y. Hou, C. Liu, Y. Liu, “Stereo matching algorithm based on improved Census transform and texture filtering”, Optik, vol. 249, 2021, p. 168186, doi: 10.1016/J.IJLEO.2021.168186.
    [6] A. F. Kadmin, R. A. Hamzah, T. F. T. Wook, “Local stereo matching algorithm using modified dynamic cost computation”, Indonesian Journal of Electrical Engineering and Computer Science, vol. 22, no. 3, 2021, pp. 1312–1319. doi: 10.11591/IJEECS.V22.I3.PP1312-1319.
    [7] Li Z., Huang J., Wang W., & Huang Y. (2024). A new stereo matching algorithm based on improved four-moded census transform and adaptive cross pyramid model. Electronic Research Archive, vol. 32, no. 7, 4340–4364. doi: 10.3934/era.2024195.
    [8] D. Wei, Y. Zhang, C. Li, “Robust line segment matching via reweighted random walks on the homography graph,” Pattern Recognit, vol. 111, 2021, p. 107693, doi: 10.1016/J.PATCOG.2020.107693.
    [9] H. Liu, “Stereo matching algorithm based on two-phase adaptive optimization of AD-census and gradient fusion”, IEEE International Conference on Real-Time Computing and Robotics, 2021, pp. 726–731, doi: 10.1109/RCAR52367.2021.9517511
    [10] D. Xi, H. Yang, B. Tan,” Stereo matching algorithm based on improved census transform and minimum spanning tree cost aggregation”, J Vis Commun Image Represent, vol. 98, 2023, pp. 104023-104038, doi: 10.1016/J.JVCIR.2023.104023.
    [11] S. Yang, X. Lei, Z. Liu, G.Sui,” An efficient local stereo matching method based on an adaptive exponentially weighted moving average filter in SLIC space”, IET Image Process, vol. 15, no. 8, 2021, pp. 1722–1732. doi: 10.1049/IPR2.12140.
    [12] B. Hou, H. Ma, Z. Li, W. Zhang,” Segment-tree based stereo matching algorithm of segmented regions collaborative optimization”, International Conference on Electronic Information Engineering, 2022, pp. 209-216, doi: 10.1117/12.2634688.
    [13] Z. Huang, G. Wan, X. Tao, “Adaptive Local Stereo Matching Based on Improved Census Transform”, Proceedings of China Automation Congress, 2023, pp. 2144–2149, doi: 10.1109/CAC59555.2023.10451091.
    [14] A. F. Kadmin, R. A. Hamzah, M. N. Abd Manap, S. F. Abd Gani,” Improved stereo matching algorithm based on census transform and dynamic histogram cost computation”, International Journal of Emerging Technology and Advanced Engineering, vol. 11, no. 8, 2021, pp 48–57, doi: 10.46338/IJETAE0821_07.
    [15] L. Zhao, F. Guo, B. Zhou, “A Generalized Voronoi Diagram-Based Segment-Point Cyclic Line Segment Matching Method for Stereo Satellite Images”, Remote Sensing, vol. 16, no. 23, 2024, p. 4395, doi: 10.3390/RS16234395.
    [16] J. Zhang, Y. Zhang, C. Wang, C. Qin, “Binocular stereo matching algorithm based on MST cost aggregation”, Mathematical Biosciences and Engineering, vol. 18, no. 4, 2021, pp. 3215–3226. doi: 10.3934/MBE.2021160.
    [17] X. Li, Y. Bai, “Stereo Matching Algorithm Based on Improved Census Transform and Fusion Cost”, International Conference on Communication Technology and Information Technology, 2023, pp. 176–180, doi: 10.1109/ICCTIT60726.2023.10435961
    [18] H. Wei, L. Meng, “An accurate stereo matching method based on color segments and edges,” Pattern Recognit, vol. 133, 2023, p. 108996. doi: 10.1016/J.PATCOG.2022.108996
    [19] C. Braunstein, E. Ilg, V. Golyanik, “Quantum-Hybrid Stereo Matching with Nonlinear Regularization and Spatial Pyramids”, International Conference on 3D Vision, 2024, pp. 1340–1349, doi: 10.1109/3DV62453.2024.00121.
    [20] X. Chen, W. Zhang, Y. Hou, L. Yang, “Improved stereo matching algorithm based on multi-scale fusion”, Journal of Northwestern Polytechnical University, vol. 39, no. 4, 2021, pp. 876–882. doi: 10.1051/JNWPU/20213940876.
    [21] G. Yao, “Quasi-Dense Matching for Oblique Stereo Images through Semantic Segmentation and Local Feature Enhancement”, Remote Sensing, vol. 16, no. 4, 2024, p. 632, doi: 10.3390/RS16040632.
    [22] S. Wu, X. Wang, J. He, X. Song, “Research on Trans. Line Stereo Matching Based on Twin Residual Network”, Conf. on Electronics and Electrical Eng. Tech., 2022, pp. 90–97, doi: 10.1109/EEET58130.2022.00024.
    [23] K. Yit Kok, P. Rajendran, ”A local multi-block with various constraints approach for stereo vision in StereoPi”, Measurement, vol. 200, 2022, p. 111617, doi: 10.1016/J.MEASUREMENT.2022.111617
    [24] N. Liu, N. Zhao, X. Ouyang, “Context Geometry Volume and Warping Refinement for Real-Time Stereo Matching”, Electronics, vol. 14, no. 5, 2025, p. 892, doi: 10.3390/ELECTRONICS14050892.
    [25] Z. Xu, J. Wang, Y. Wang, “A Dual Branch Multiscale YStereo Matching Network for High-Resolution Satellite Remote Sensing Images”, IEEE J Sel Top Appl Earth Obs Remote Sens, vol. 1v, 2025, pp. 949–964, doi: 10.1109/JSTARS.2024.3502842.
    [26] L. Dong, Y. Han, M. Hu, Y. Wang, ”Stereo matching method based on high-precision occlusion-recovering and discontinuity-preserving”, Displays, vol. 80, 2023, p. 102573, doi: 10.1016/J.DISPLA.2023.102573.
    [27] D. Scharstein, et al., “High-Resolution Stereo Datasets with Subpixel-Accurate Ground Truth”, Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics, 2014, pp. 31–42, doi: 10.1007/978-3-319-11752-2_3/COVER.
    [28] L. Kong, J. Zhu, S. Ying, ”Stereo Matching Based on Guidance Image and Adaptive Support Region”, Acta Optica Sinica, vol. 40, no. 9, 2020, p. 0915001, doi: 10.3788/AOS202040.0915001.
    [29] H. Yaoyu, Z. Weikun, S. Sebastian, “Deep-Learning Assisted High-Resolution Binocular Stereo Depth Reconstruction”, IEEE International Conference on Robotics and Automation, 2020, pp. 8637–8643.
    [30] S. Fan, W. Sun, J. Zheng, W. Wu, “Accurate edge-preserving stereo matching by enhancing anisotropy”, Signal Process Image Commun, vol. 114, 2023, p. 116945, doi: 10.1016/J.IMAGE.2023.116945.
    [31] Y. Fu, K. Lai, W. Chen, Y. Xiang, “A pixel pair–based encoding pattern for stereo matching via an adaptively weighted cost”, IET Image Process, vol. 15, no. 4, 2021, pp. 908–917, doi: doi.org/10.1049/ipr2.12071.
    [32] Y. Zhang, Q. Shen, P. Song.” Local Side Window Algorithm with Tree Segmentation for Stereo Matching”, Laser & Optoelectronics Progress, vol. 61, no. 22, 2021, pp. 2237007-2237007, doi: 10.3788/LOP240703.
    [33] B. Ruf, J. Mohrs, M. Weinmann, J. Beyerer, “ReS2tAC—UAV-Borne Real-Time SGM Stereo Optimized for Embedded ARM and CUDA Devices”, Sensors, Vol. 21, No. 11, 2021, p. 3938, doi: 10.3390/S21113938
    [34] J. Min, J. Kim, C. H Min, M. Choi,” DepthFocus: Controllable Depth Estimation for See-Through Scenes”, arXiv preprint arXiv:2511.16993., 2025, p. 1. doi.org/10.48550/arXiv:2511.16993
    [35] Y. Wang, K. Li, L. Wang, J. Hu, D. O. Wu, and Y. Guo, “Adstereo: Efficient stereo matching with adaptive downsampling and disparity alignment”, IEEE Transactions on Image Processing, vol. 34, 2025, pp. 1204-1218. doi.org/10.1109/TIP.2025.3540282.
    [36] Z. Congxuan, W. Junjie, J. Shaofeng, “Dense-CNN: Dense convolutional neural network for stereo matching using multiscale feature connection”, Signal Processing: Image Communication, vol. 95, 2021, p. 116285. doi.org/10.1016/j.image.2021.116285.
    [37] A. Gomez Chavez, A. Ranieri, D. Chiarella and A. Birk, “Underwater Vision-Based Gesture Recognition: A Robustness Validation for Safe Human–Robot Interaction”, IEEE Robotics & Automation Magazine, vol. 28, no. 3, pp. 67-78. doi.org/10.1109/MRA.2021.3075560
    [38] G., Li, S. Huang, Z. Yin, J. Li, and K. Zhang, “Underwater Refractive Stereo Vision Measurement and Simulation Imaging Model Based on Optical Path”, Journal of Marine Science and Engineering, vol. 12, no. 11, 2024, p. 1955. doi.org/10.3390/jmse12111955.
    [39] L. Zhu, Y. Gao, J. Zhang, Y. Li, and X. Li,” Reliable and Effective Stereo Matching for Underwater Scenes. Remote Sensing”, vol. 16, no. 23, 2024, p. 4570. doi.org/10.3390/rs16234570
    [40] Q. Lv, J. Dong, Y. Li, S. Chen, and W. Wang, “UWStereo: A Large Synthetic Dataset for Underwater Stereo Matching”, arXiv preprint, arXiv:2409.01782. 2025. doi.org/10.48550/arxiv.2409.01782.

Nurliyana Abd Mutalib, Rostam Affendi Hamzah, Ken Prameswari Caesarella Aryaputri, Ahmad Fauzan Kadmin, Masrullizam Mat Ibrahim, and Adi Irwan Herman (2026), Depth Map Reconstruction in Underwater Robotic Vision using Gradient Spanning Trees and Boundary Segmentation. IJEER 14(3), 822-831. DOI: 10.37391/IJEER.140316.