Abstract

Image stitching technology can synthesize multiple images with overlapping areas into a panorama, allowing users to acquire a more complete field of view. However, mainstream image warping methods, when processing images with significant parallax, apply differential deformation operations to pixels on straight-line structures, causing originally straight lines to become bent, broken, or misaligned in the stitched result. This leads to visual distortion, a lack of realism, reduced stitching quality and usability, and amplifies errors in subsequent processing. To address this, this paper proposes a parallax-tolerant image stitching algorithm with straight-line structure preservation. Firstly, the SOLD2 line detection method is employed to obtain line structures and acquire line endpoints and sample points, providing data for point collinearity constraints and global line construction. Secondly, global lines are constructed using the obtained line structures through parallelism constraints, collinearity constraints, and endpoint distance constraints, providing data for point collinearity constraints to ensure the correct distribution of line structures in the image. Finally, point collinearity constraints based on line endpoints and sample points are added to the REW method to enhance line protection. Experimental results show that the proposed algorithm mitigates the issue of straight-line structure distortion in stitched results from images with large parallax faced by mainstream algorithms. The RMSE on multiple datasets is on average about 0.068 lower than the LPC method and significantly better than other compared algorithms.

References

• Ghosh, D., & Kaabouch, N. (2016). A survey on image mosaicing techniques. Journal of Visual Communication and Image Representation, 34,1–11.
• Wang, J., Wang, C., Chen, W., Dou, Q., & Chi, W. (2024). Embracing the future: The rise of humanoid robots and embodied AI. Intelligent Robotics, 4,196–209.
• Chen, X., et al. (2025). Towards intelligent shipping: Image-enhanced ship detection and situation analysis in low-light scenes. Intelligent Robotics, 5(3), 662–678.
• Zhang, W., Wang, Y., & Liu, Y. (2022). Generating high-quality panorama by view synthesis based on optical flow estimation. Sensors, 22(2), 470.
• Wang, J., Du, P., Yang, S., Zhang, Z., & Ning, J. (2024). A spatial arrangement preservation-based stitching method via geographic coordinates of UAV for farmland remote sensing image. IEEE Transactions on Geoscience and Remote Sensing, 62,1–13.
• Lu, T., Li, S., & Fu, W. (2014). Fusion based seamless mosaic for remote sensing images. Sensing and Imaging, 15,101.
• Bai, Z., et al. (2020). Real-time video stitching for mine surveillance using a hybrid image registration method. Electronics, 9(9), 1336.
• Ha, H. G., Jung, K., Lee, S., et al. (2021). Heterogeneous stitching of X-ray images according to homographic evaluation. Journal of Digital Imaging, 34,1249–1263.
• Lowe, D. G. (2004). Sift-the scale invariant feature transform. International Journal of Computer Vision, 2(91-110), 2.
• Brown, M., & Lowe, D. G. (2007). Automatic panoramic image stitching using invariant features. International Journal of Computer Vision, 74(1), 59–73.
• Huang, J.-B., Kang, S. B., & Ahuja, N. (2013). As-projective-as-possible image stitching with moving DLT. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)(pp. 2339–2346).
• Jia, Q., Li, Z., Fan, X., Zhao, H., Teng, S., Ye, X., & Latecki, L. J. (2021). Leveraging line-point consistence to preserve structures for wide parallax image stitching. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)(pp. 1–10).
• Pautrat, R., Lin, J.-T., Larsson, V., Oswald, M. R., & Pollefeys, M. (2021). SOLD2: Self-supervised occlusion-aware line description and detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)(pp. 11368–11378).
• Li, J., Wang, Z., Lai, S., et al. (2018). Parallax-tolerant image stitching based on robust elastic warping. IEEE Transactions on Multimedia, 20(7), 1672–1687.
• Yang, L., Tian, B., & Dang, J. (2024). Survey on image stitching algorithm based on deep learning. Application Research of Computers, 41(7), 1930–1939.