Download

Automated object recognition systems on complex background require high-quality, high-resolution images of the analyzed scenes. The natural conditions of the scene are an important factor of image acquisition. It may be characterized by the excessive lighting from point sources, accompanied by the presence of shadows. At the same time, in the tasks of moving objects recognition, the presence of shadows in the image contributes to the merging of these objects and therefore complicates their recognition. Thus, the solution of the problem of shadow removal from images remains a constant subject of research, with the aim of developing computationally simple and effective algorithms. A brief analysis of existing algorithms for removing shadows from images is proposed. The following basic approaches to solving this problem are highlighted: the use of segmentation to highlight shadows; the construction, formation, and solution of variational problems; the use of neural networks, and the use of SSR (Single Scale Retinex) technology. The research is based on SSR technology, which was pioneered by E. Land. The novelty of proposed method is that the normalized function for evaluating the similarity of pixel intensity values and the smoothed value of its neighborhood has a symmetrical shape relative to its center. A comparison of the proposed technology with one described by McCann was carried out and illustrated with the results of test image processing. Thanks to the symmetry of the proposed method, better shadow removal is achieved in areas with high intensity values. The development of this symmetric approach in the context of Multi Scale Retinex development may be the subject of further research.

1. Dorrani, Z.; Farsi, H.; Mohamadzadeh, S. Shadow Removal in Vehicle Detection Using ResUNet-a. Iranian Journal of Energy and Environment 2023, 14, 87-95. https://doi.org/10.5829/IJEE.2023.14.01.11

2. Umair Arif, M.; Farooq, M.U.; Raza, R.H.; Lodhi, Z.U.A.; Hashmi, M.A.R. Comprehensive Review of Vehicle Detection Techniques Under Varying Moving Cast Shadow Conditions Using Computer Vision and Deep Learning. IEEE Access 2022, 10, 104863-104886. https://doi.org/10.1109/ACCESS.2022.3208568

3. Karnieli, A.; Fried, O.; Hel-Or, Y. DeepShadow: Neural Shape from Shadow. In: Avidan, S., Brostow, G., Cisse, M., Farinella, G.M., Hassner, T. (eds) Computer Vision - ECCV 2022. Lecture Notes in Computer Science, 13662. Springer, Cham. https://doi.org/10.1007/978-3-031-20086-1_24

4. Rasheed, M.T.; Guo, G.; Shi, D.; Khan, H.; Cheng, X. An Empirical Study on Retinex Methods for Low-Light Image Enhancement. Remote Sens. 2022, 14, 4608. https://doi.org/10.3390/rs14184608

5. Kimmel, R.; Elad, M.; Shaked, D.; Keshet, R.; Sobel, I. A variational framework for retinex. International Journal of Computer Vision 2003, 52 (1), 7-23. https://doi.org/10.1023/A:1022314423998

6. Sun, J.; Tian, J.; Du, Y.; Tang, Y. Retinex theory-based shadow detection and removal in single outdoor image. Industrial Robot: An International Journal 2009, 36 (3), 263-269. https://doi.org/10.1108/01439910910950531

7. Li, M.; Liu, J.; Yang, W.; Sun, X.; Guo, Z. Structure-revealing low-light image enhancement via robust retinex model. IEEE Trans Image Process. 2018, 27, 2828-2841. https://doi.org/10.1109/TIP.2018.2810539

8. Fan, H.; Han, M.; Li, J. Image Shadow Removal Using End-To-End Deep Convolutional Neural Networks. Appl. Sci. 2019, 9, 1009. https://doi.org/10.3390/app9051009

9. Guo, L.; Wang, C.; Wang, Y.; Yu, Y.; Huang, S.; Yang, W.; Kot, A. C.; Wen, B. Single-Image Shadow Removal Using Deep Learning: A Comprehensive Survey. ArXiv. 2024. https://arxiv.org/abs/2407.08865.

10. McCann, J. Retinex at 50: color theory and spatial algorithms, a review. J. Electron. Imag. 2017, 26 (3), 031204. https://doi.org/10.1117/1.JEI.26.3.031204

11. Jourlin, M.; Pinoli, J.-C. A Model for Logarithmic Image Processing. Journal of Microscopy 1988, 149 (1), 21-35. https://doi.org/10.1111/j.1365-2818.1988.tb04559.x

12. Patrascu, V.; Buzuloiu, V. A Mathematical Model for Logarithmic Image Processing. Proceedings of the 5th World Multi-Conference on Systemics, Cybernetics and Informatics, Orlando, USA, July 22-25, 2001; 13, pp 117-122.

13. Berehulyak, O.; Vorobel, R. The Algebraic Model with an Asymmetric Characteristic of Logarithmic Transformation, Proceedings of the 2020 IEEE 15th International Conference on Computer Sciences and Information Technologies (CSIT), Zbarazh, Ukraine, Sept. 23-26, 2020; pp 119-122. https://doi.org/10.1109/CSIT49958.2020.9321906

14. Vorobel, R. Logarithmic Image Processing; Naukova Dumka: Kyiv, 2012. (In Ukrainian)

15. Funt, B.; Ciurea F.; McCann, J. Retinex in Matlab. J. Electron. Imag. 2004, 13, 48-57.

16. Vorobel, R.A.; Berehulyak, O.R.; Ivasenko, I.B.; Mandziy, T.S. Local image contrast enhancement based on power model of visual image perception. Information Extraction and Processing 2024, 52 (128), 95-106. https://doi.org/10.15407/vidbir2024.52.095