Software method for image segmentation under variable lighting conditions
DOI:
https://doi.org/10.34185/1562-9945-3-164-2026-08Keywords:
semantic segmentation, variable lighting, image processing, neural networks, software engineeringAbstract
This paper addresses the problem of image segmentation under variable lighting conditions, which is a critical factor in computer vision tasks. Various approaches to image processing are analyzed, including classical methods and modern neural network models. It has been established that traditional approaches have limitations due to their dependence on low-level features, whereas neural network methods demonstrate reduced performance under unstable lighting conditions.
The purpose of the research is to develop an image segmentation method that ensures adaptation to lighting conditions and improves the accuracy of object detection. A software method is proposed that combines an illumination level classification, adaptive image preprocessing, and neural network segmentation. Images are classified based on an analysis of statistical characteristics of relative luminance, which allows determining the processing algorithm for each image.
The effectiveness of the method was evaluated through experimental research using the U-Net and SegNet neural network models. A set of images with simulated lighting conditions was used for the analysis. As a result of the experiments, the proposed method demonstrated improvements of 4.95% and 1.5%, respectively, in terms of the mIoU and Pixel Accuracy metrics for the SegNet model, as well as 1.6% and 0.07% for the U-Net model. A comparative analysis has shown that the use of adaptive image preprocessing improves segmentation quality regardless of the chosen neural network architecture.
The proposed method increases the robustness of segmentation systems to lighting changes, reduces the impact of noise and distortions, and ensures more stable model performance in real-world conditions. This makes it an effective solution for a wide range of computer vision tasks, particularly in systems operating in dynamic environments.
References
Anwesh, K., Pal, D., Ganguly, D., Chatterjee, K., & Roy, S. (2022). Number plate recogni-tion from enhanced super-resolution using generative adversarial network. Multimedia Tools and Applications, 1–17. https://doi.org/10.1007/s11042-022-14018-0
Kinahan, J., & Smeaton, A. F. (2021). Image segmentation to identify safe landing zones for unmanned aerial vehicles. arXiv. https://doi.org/10.48550/arXiv.2111.14557
Jin, B., Cruz, L., & Gonçalves, N. (2020). Deep facial diagnosis: Deep transfer learning from face recognition to facial diagnosis. IEEE Access, 8, 123649–123661. https://doi.org/10.1109/ACCESS.2020.3005687
Zhang, L. (2008). In situ image segmentation using the convexity of illumination distribu-tion of the light sources. IEEE Transactions on Pattern Analysis and Machine Intelligence, 30(10), 1786–1799. https://doi.org/10.1109/TPAMI.2007.70830
Abosinnee, A. S., Bencsik, G., & Abedi, F. (2025). Edges in image with illumination varia-tions scenarios: A review. The Visual Computer, 41, 12277–12305. https://doi.org/10.1007/s00371-025-04157-4
Chen, Y., Wen, C., Liu, W., & Li, G. (2023). A depth iterative illumination estimation network for low-light image enhancement based on Retinex theory. Scientific Reports, 13(1), 19709. https://doi.org/10.1038/s41598-023-46693-w
Sun, Z., Zhu, H., Xiao, X., Gu, Y., & Xu, Y. (2024). Nighttime image semantic segmenta-tion with Retinex theory. Image and Vision Computing, 148, 105149. https://doi.org/10.1016/j.imavis.2024.105149
Vrshnak, D., Domislović, I., Subašić, M., & Lončarić, S. (2022). Illuminant segmentation for multi-illuminant scenes using latent illumination encoding. Signal Processing: Image Communication, 108, 116822. https://doi.org/10.1016/j.image.2022.116822
Nugraha, D. G., Purwanto, D., Dikairono, R., Widjiati, E., & Adinanta, H. (2025). Com-parative Analysis of Image Segmentation Methods for Unmanned Surface Vehicles under Varying Illumination Conditions. In 26th International Seminar on Intelligent Technology and Its Applications: Fostering Equal Opportunities for Breakthrough Technology Innovations, ISITIA 2025 - Proceedings (2025 ed., pp. 88-93). Institute of Electrical and Electronics Engi-neers Inc.. https://doi.org/10.1109/ISITIA66279.2025.11137505
Cardenas-Gallegos, J. S., Severns, P. M., Klimeš, P., Lacerda, L. N., Peduzzi, A., & Fer-rarezi, R. S. (2025). Reliable plant segmentation under variable greenhouse illumination con-ditions. Computers and Electronics in Agriculture, 229, 109711. https://doi.org/10.1016/j.compag.2024.109711
Ronneberger, O., Fischer, P., & Brox, T. (2015). U-Net: Convolutional networks for bi-omedical image segmentation. https://doi.org/10.48550/arXiv.1505.04597
Badrinarayanan, V., Kendall, A., & Cipolla, R. (2017). SegNet: A deep convolutional encoder-decoder architecture for image segmentation. IEEE TPAMI. https://doi.org/10.1109/TPAMI.2016.2644615
Intelec AI. (2019). Car segmentation [Data set]. Kaggle. https://www.kaggle.com/datasets/intelecai/car-segmentation
Downloads
Published
Issue
Section
License
Copyright (c) 2026 System technologies
This work is licensed under a Creative Commons Attribution 4.0 International License.
