Predicting Filtered Image Quality Using Transfer Learning on Sentinel-1 Speckle Noise with DenseNet-121
DOI:
https://doi.org/10.36023/ujrs.2025.12.4.293Keywords:
Speckle noise, Sentinel-1, image quality metrics, DenseNet-121, transfer learningAbstract
Speckle noise inherent to synthetic aperture radar (SAR) imagery degrades image quality and complicates automated analysis in Earth observation applications. Quantitative assessment of despeckling results requires computing quality metrics against reference images, which are unavailable in operational SAR scenarios. This paper presents a method for a priori prediction of filtered Sentinel-1 SAR image quality metrics before applying speckle noise filters. Unlike existing approaches predicting relative quality improvement, the proposed method predicts absolute values of five metrics (PSNR, WSNR, SSIM, MS-SSIM, FSIM) for a specific filter, enabling direct comparison and rational filter selection. The methodology employs transfer learning of DenseNet-121 convolutional neural network, pre-trained on ImageNet, adapted for single-channel SAR inputs through architectural modifications including input layer transformation, pooling optimization, and regression head replacement. A novel synthetic data generation pipeline utilizes histogram matching of Sentinel-2 optical images with Sentinel-1 SAR references to create training samples preserving ground truth. Dynamic gamma-distributed speckle noise addition with variable ENL ∈ [2, 6] enhances data variability and model robustness. Experiments with six classical filters (Gamma MAP, Lee, Enhanced Lee, Frost, SRAD, Kuan) demonstrate high prediction accuracy across all filter-metric combinations. The coefficient of determination R² reaches 0.997 for best combinations and exceeds 0.97 for most of the 30 trained models. Mean absolute prediction errors remain below 0.29 dB for PSNR and 0.014 for SSIM across all tested configurations. The approach enables a priori quality prediction without reference images, allowing optimization of SAR processing workflows and resource planning before resource-intensive despeckling.
Contributions of Authors: Conceptualization, Raed A. and Oleksii R.; methodology, Raed A.; formal analysis, Raed A.; investigation, Raed A. and Oleksii R.; data curation, Raed A. and Oleksii R.; writing—original draft preparation, Raed A.; writing—review and editing, Raed A. and Oleksii R.; visualization, Raed A. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Data Availability Statement: The processed dataset is available in the Zenodo repository (https://zenodo.org/uploads/17253925). Python implementation code is available in the GitHub repository (https://github.com/rsenaikh/Predicting_Quality_after_Noise_Removal).
Acknowledgments: The authors would like to express their sincere gratitude to the Copernicus Data Space Ecosystem for providing open access to Sentinel-1 and Sentinel-2 data. We are also grateful to reviewers and editors for their valuable comments, recommendations, and attention to the work.
Conflicts of Interest: The authors declare no conflict of interest.
References
Abramov, S., Krivenko, S., Roenko, A., & Lukin, V. (2013). Prediction of filtering efficiency for DCT-based image denoising. 2nd Mediterranean Conference on Embedded Computing (MECO). https://doi.org/10.1109/MECO.2013.6601327
Beauchemin, M., Thomson, K. P. B., & Edwards, G. (1996). Optimization of the Gamma-Gamma MAP filter for SAR image clutters. International Journal of Remote Sensing, 17(5), 1063-1067. https://doi.org/10.1080/01431169608949067
European Space Agency. (2016). Sentinel-1 Product Definition (S1-RS-MDA-52-7440, Ed. 2/7).
European Space Agency. (2022). Sentinel-1 Level 1 Detailed Algorithm Definition (Ed. 2.5).
Filipponi, F. (2019). Sentinel-1 GRD preprocessing workflow. Proceedings, 18(11). https://doi.org/10.3390/ECRS-3-06201
Frost, V. S., Stiles, J. A., Shanmugan, K. S., & Holtzman, J. C. (1982). A model for radar images and its application to adaptive digital filtering of multiplicative noise. IEEE Transactions on Pattern Analysis and Machine Intelligence, 4(2), 157-166. https://doi.org/10.1109/TPAMI.1982.4767223
IEEE Standard for the Deep Learning-Based Assessment of Visual Experience Based on Human Factors. (2022). IEEE Std 3333.1.3-2022. https://doi.org/10.1109/IEEESTD.2022.9781357
Kuan, D. T., Sawchuk, A. A., Strand, T. C., & Chavel, P. (1985). Adaptive noise smoothing filter for images with signal-dependent noise. IEEE Transactions on Pattern Analysis and Machine Intelligence, 7(2), 165-177. https://doi.org/10.1109/TPAMI.1985.4767641
Lee, J. S. (1980). Digital image enhancement and noise filtering by use of local statistics. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2(2), 165-168. https://doi.org/10.1109/TPAMI.1980.4766994
Lukin, V., Abramov, S., Rubel, A., Krivenko, S., Naumenko, A., Vozel, B., Chehdi, K., Egiazarian, K., & Astola, J. (2014). An approach to prediction of signal-dependent noise removal efficiency by DCT-based filter. Telecommunications and Radio Engineering, 73(18), 1645-1659. https://doi.org/10.1615/TelecomRadEng.v73.i18.40
Medeiros, F. N. S., Mascarenhas, N. D. A., & Costa, L. F. (2003). Evaluation of speckle noise MAP filtering algorithms applied to SAR images. International Journal of Remote Sensing, 24(24), 5197-5218. https://doi.org/10.1080/0143116031000115148
Moein, F., & Taban, M. R. (2024). Removing speckle noise from synthetic-aperture radar images using artificial intelligence techniques. 2024 11th International Symposium on Telecommunications (IST). https://doi.org/10.1109/IST64061.2024.10843539
Rubel, O., Lukin, V., & de Medeiros, F. (2015). Prediction of despeckling efficiency of DCT-based filters applied to SAR images. Proc. IEEE International Conference on Distributed Computing in Sensor Systems (DCOSS), 159-168. https://doi.org/10.1109/DCOSS.2015.16
Rubel, O., Lukin, V., Rubel, A., & Egiazarian, K. (2019). NN-based prediction of Sentinel-1 SAR image filtering efficiency. Geosciences, 9(7), 290. https://doi.org/10.3390/geosciences9070290
Rubel, O., Lukin, V., Rubel, A., & Egiazarian, K. (2020). Prediction of Lee filter performance for Sentinel-1 SAR images. Electronic Imaging, 2020(9), Article IQSP-371. https://doi.org/10.2352/ISSN.2470-1173.2020.9.IQSP-371
Rubel, O., Lukin, V., Rubel, A., & Egiazarian, K. (2021). Selection of Lee filter window size based on despeckling efficiency prediction for Sentinel SAR images. Remote Sensing, 13, 1887. https://doi.org/10.3390/rs13101887
Singh, P., & Pandey, R. (2016). Speckle noise: Modelling and implementation. International Science Press, 9(17), 8717-8727.
Sveinsson, J. R., & Benediktsson, J. A. (1996). Speckle reduction and enhancement of SAR images in the wavelet domain. IGARSS ’96. 1996 International Geoscience and Remote Sensing Symposium, 36. doi:https://doi.org/10.1109/IGARSS.1996.516245
Talebi, H., & Milanfar, P. (2018). NIMA: Neural image assessment. IEEE Transactions on Image Processing, 27(8), 3998-4011. https://doi.org/10.1109/TIP.2018.2831899
Wanjiku, R. N., Nderu, L., & Kimwele, M. W. (2022). Transfer learning data adaptation using conflation of low‐level textural features. Engineering Reports. https://doi.org/10.1002/eng2.12603
Wang, Z., Bovik, A. C., Sheikh, H. R., & Simoncelli, E. P. (2004). Image quality assessment: from error visibility to structural similarity. IEEE Transactions on Image Processing, 13(4), 600-612. https://doi.org/10.1109/TIP.2003.819861
Wei, S., Zhang, H., Zeng, X., Zhou, Z., Shi, J., & Zhang, X. (2022). CARNet: An effective method for SAR image interference suppression. International Journal of Applied Earth Observation and Geoinformation, 114, 103019. https://doi.org/10.1016/j.jag.2022.103019
Yu, Y., & Acton, S. T. (2002). Speckle reducing anisotropic diffusion. IEEE Transactions on Image Processing, 11(11), 1260-1270. https://doi.org/10.1109/TIP.2002.804276
Zhang, L., Zhang, L., Mou, X., & Zhang, D. (2011). FSIM: A feature similarity index for image quality assessment. IEEE Transactions on Image Processing, 20(8), 2378-2386. https://doi.org/10.1109/TIP.2011.2109730
Downloads
Published
How to Cite
Issue
Section
License
Licensing conditions: the authors retain their copyrights and grant the journal the right of first publication of a work, simultaneously licensed in accordance with the Creative Commons Attribution License International CC-BY, which allows you to share the work with proof of authorship of the work and initial publication in this journal.
The authors, directing the manuscript to the editorial office of the Ukrainian Journal of Remote Sensing of the Earth, agree that the editorial board transfers the rights to protection and use of the manuscript (material submitted to the journal editorial board, including such protected copyright objects as photographs of the author, drawings, charts, tables, etc.), including reproduction in print and on the Internet; for distribution; to translate the manuscript into any languages; export and import of copies of the journal with the article of the authors for the purpose of distribution, informing the public. The above rights are transferred by the authors to the editors, without limitation of their validity, and in the territory of all countries of the world without limitation, including in Ukraine.
The authors guarantee that they have exclusive rights to use the submitted material. The editors are not liable to third parties for breach of data by the authors of the guarantees. The authors retain the right to use the published material, its fragments and parts for personal, including scientific and educational purposes. The rights to the manuscript are considered to be transferred by the authors of the editorial board from the moment of the publication of the issue of the journal in which it is published. Reprinting of materials published in the journal by other individuals and legal entities is possible only with the consent of the publisher, with the obligatory indication of the issue of the journal in which the material was published.
