Generative Adversarial Network-Based Lightweight High-Dynamic-Range Image Reconstruction Model
Source abstract
Abstract
The generation of High-Dynamic-Range (HDR) images is essential for capturing details at various brightness levels, but current reconstruction methods, using deep learning techniques, often require significant computational resources, limiting their applicability on devices with moderate resources. In this context, this paper presents a lightweight architecture for reconstructing HDR images from three Low-Dynamic-Range inputs. The proposed model is based on Generative Adversarial Networks and replaces traditional convolutions with depthwise separable convolutions, reducing the number of parameters while maintaining high visual quality and minimizing luminance artifacts. The evaluation of the proposal is conducted through quantitative, qualitative, and computational cost analyses based on the number of parameters and FLOPs. Regarding the qualitative analysis, a comparison between the models was performed using samples that present reconstruction challenges. The proposed model achieves a PSNR-μ of 43.51 dB and SSIM-μ of 0.9917, achieving competitive quality metrics comparable to HDR-GAN while reducing the computational cost by 6× in FLOPs and 7× in the number of parameters, using approximately half the GPU memory consumption, demonstrating an effective balance between visual fidelity and efficiency.
Source: crossref · Quality: baseline
Research context
OpenAlex topics: Advanced Image Processing Techniques · Image and Signal Denoising Methods · Image Processing Techniques and Applications
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
Affiliations
PAVIC Laboratory, University of Acre (UFAC), Rio Branco 69915-900, Brazil
Funding
PAVIC Laboratory
References (49)
Show source-provided references
1. High dynamic range imaging—A review
2. High dynamic range display systems
3. Deep learning for hdr imaging: State-of-the-art and future trends10.1109/TPAMI.2021.3123686 ↗
4. A review on high-dynamic-range imaging with its technique
5. Demystifying high-dynamic-range technology: A new evolution in digital media10.1109/MCE.2015.2463294 ↗
6. Xu, F., Liu, J., Song, Y., Sun, H., and Wang, X. (2022). Multi-exposure image fusion techniques: A comprehensive review. Remote Sens., 14.10.3390/rs14030771 ↗
7. Recovering high dynamic range radiance maps from photographs
8. Prabhakar, K.R., Agrawal, S., Singh, D.K., Ashwath, B., and Babu, R.V. (2020, January 23–28). Towards practical and efficient high-resolution HDR deghosting with CNN. Proceedings of the Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK. Proceedings, Part XXI 16.10.1007/978-3-030-58589-1_30 ↗
9. Hu, J., Gallo, O., Pulli, K., and Sun, X. (2013, January 23–28). HDR deghosting: How to deal with saturation?. Proceedings of the IEEE Conference on coMputer Vision and Pattern Recognition, Portland, OR, USA.10.1109/CVPR.2013.154 ↗
10. Deep High Dynamic Range Imaging of Dynamic Scenes
11. Generative adversarial nets
12. Raipurkar, P., Pal, R., and Raman, S. (2021, January 20–22). HDR-cGAN: Single LDR to HDR image translation using conditional GAN. Proceedings of the Twelfth Indian Conference on Computer Vision, Graphics and Image Processing, Virtual Event.10.1145/3490035.3490275 ↗
13. Wang, C., Serrano, A., Pan, X., Chen, B., Seidel, H.P., Theobalt, C., Myszkowski, K., and Leimkuehler, T. (2023). GlowGAN: Unsupervised Learning of HDR Images from LDR Images in the Wild. arXiv.10.1109/ICCV51070.2023.00964 ↗
14. MEF-GAN: Multi-Exposure Image Fusion via Generative Adversarial Networks10.1109/TIP.2020.2999855 ↗
15. Uphdr-gan: Generative adversarial network for high dynamic range imaging with unpaired data10.1109/TCSVT.2022.3190057 ↗
16. Hdr-gan: Hdr image reconstruction from multi-exposed ldr images with large motions10.1109/TIP.2021.3064433 ↗
17. Ten years of generative adversarial nets (GANs): A survey of the state-of-the-art10.1088/2632-2153/ad1f77 ↗
18. Tel, S., Heyrman, B., and Ginhac, D. (2022, January 23–27). Cen-hdr: Computationally efficient neural network for real-time high dynamic range imaging. Proceedings of the European Conference on Computer Vision, Tel Aviv, Israel.10.1007/978-3-031-25063-7_23 ↗
19. Chollet, F. (2017, January 21–26). Xception: Deep Learning with Depthwise Separable Convolutions. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA.10.1109/CVPR.2017.195 ↗
20. Howard, A.G. (2017). Mobilenets: Efficient convolutional neural networks for mobile vision applications. arXiv.
21. SynthNet: A skip connected depthwise separable neural network for Novel View Synthesis of solid objects10.1016/j.rineng.2022.100383 ↗
22. Krishnan, K.S., and Krishnan, K.S. (2021, January 1–2). SwiftSRGAN-Rethinking super-resolution for efficient and real-time inference. Proceedings of the 2021 International Conference on Intelligent Cybernetics Technology & Applications (ICICyTA), Bandung, Indonesia.10.1109/ICICyTA53712.2021.9689188 ↗
23. Fabric defect detection using a one-class classification based on depthwise separable convolution autoencoder10.1088/1742-6596/2562/1/012053 ↗
24. Guo, Z., Wang, J., Chen, S., and Kang, Y. (2023, January 27–29). A Lightweight Stereo Matching Neural Network Based on Depthwise Separable Convolution. Proceedings of the 2023 IEEE International Conference on Integrated Circuits, Technologies and Applications (ICTA), Hefei, China.10.1109/ICTA60488.2023.10364258 ↗
25. Nemoto, H., Korshunov, P., Hanhart, P., and Ebrahimi, T. (2015, January 5–6). Visual attention in LDR and HDR images. Proceedings of the 9th International Workshop on Video Processing and Quality Metrics for Consumer Electronics (VPQM), Chandler, AZ, USA.
26. Robust patch-based hdr reconstruction of dynamic scenes10.1145/2366145.2366222 ↗
27. An objective deghosting quality metric for HDR images10.1111/cgf.12818 ↗
28. Tel, S., Wu, Z., Zhang, Y., Heyrman, B., Demonceaux, C., Timofte, R., and Ginhac, D. (2023, January 1–6). Alignment-free HDR Deghosting with Semantics Consistent Transformer. Proceedings of the ICCV, Paris, France.10.1109/ICCV51070.2023.01179 ↗
29. Kirk, K., and Andersen, H.J. (2006, January 4–7). Noise Characterization of Weighting Schemes for Combination of Multiple Exposures. Proceedings of the BMVC, Edinburgh, UK.
30. Granados, M., Ajdin, B., Wand, M., Theobalt, C., Seidel, H.P., and Lensch, H.P. (2010, January 13–18). Optimal HDR reconstruction with linear digital cameras. Proceedings of the 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, San Francisco, CA, USA.10.1109/CVPR.2010.5540208 ↗
31. Khan, E.A., Akyuz, A.O., and Reinhard, E. (2006, January 26–29). Ghost removal in high dynamic range images. Proceedings of the 2006 International Conference on Image Processing, Las Vegas, NA, USA.10.1109/ICIP.2006.312892 ↗
32. Reconstruction of high contrast images for dynamic scenes10.1007/s00371-011-0653-0 ↗
33. Tomaszewska, A., and Mantiuk, R. (2007). Image Registration for Multi-Exposure High Dynamic Range Image Acquisition, Vaclav Skala-UNION Agency.
34. Mirza, M., and Osindero, S. (2014). Conditional Generative Adversarial Nets. arXiv.
35. Learning a deep single image contrast enhancer from multi-exposure images10.1109/TIP.2018.2794218 ↗
36. The HDR photographic survey10.2352/CIC.2007.15.1.art00044 ↗
37. Kaiser, L., Gomez, A.N., and Chollet, F. (May, January 30). Depthwise Separable Convolutions for Neural Machine Translation. Proceedings of the International Conference on Learning Representations, Vancouver, BC, Canada.
38. Guo, Y., Li, Y., Wang, L., and Rosing, T. (February, January 27). Depthwise convolution is all you need for learning multiple visual domains. Proceedings of the AAAI Conference on Artificial Intelligence, Honolulu, HI, USA.
39. Navab, N., Hornegger, J., Wells, W.M., and Frangi, A.F. (2015, January 5–9). U-Net: Convolutional Networks for Biomedical Image Segmentation. Proceedings of the Medical Image Computing and Computer-Assisted Intervention—MICCAI 2015, Munich, Germany.10.1007/978-3-319-24571-3 ↗
40. Isola, P., Zhu, J.Y., Zhou, T., and Efros, A.A. (2017, January 21–26). Image-to-image translation with conditional adversarial networks. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, HI, USA.10.1109/CVPR.2017.632 ↗
41. Park, S.W., and Kwon, J. (2019, January 15–20). Sphere generative adversarial network based on geometric moment matching. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Long Beach, CA, USA.10.1109/CVPR.2019.00442 ↗
42. Scope of validity of PSNR in image/video quality assessment10.1049/el:20080522 ↗
43. Image quality assessment through FSIM, SSIM, MSE and PSNR—a comparative study10.4236/jcc.2019.73002 ↗
44. Deng, C., Ma, L., Lin, W., and Ngan, K.N. (2015). Quality Assessment in Computer Graphics. Visual Signal Quality Assessment: Quality of Experience (QoE), Springer International Publishing.10.1007/978-3-319-10368-6 ↗
45. Mantiuk, R., Kim, K.J., Rempel, A.G., and Heidrich, W. (2011, January 12–15). HDR-VDP-2: A calibrated visual metric for visibility and quality predictions in all luminance conditions. Proceedings of the ACM SIGGRAPH, Hong Kong, China.10.1145/1964921.1964935 ↗
46. HDR-VDP-2.2: A calibrated method for objective quality prediction of high-dynamic range and standard images10.1117/1.JEI.24.1.010501 ↗
47. Mantiuk, R.K., Hammou, D., and Hanji, P. (2023). HDR-VDP-3: A multi-metric for predicting image differences, quality and contrast distortions in high dynamic range and regular content. arXiv.
48. A Feature-Enriched Completely Blind Image Quality Evaluator10.1109/TIP.2015.2426416 ↗
49. Lopez-Cabrejos, J., Paixão, T., Alvarez, A.B., and Luque, D.B. (2025). An Efficient and Low-Complexity Transformer-Based Deep Learning Framework for High-Dynamic-Range Image Reconstruction. Sensors, 25.10.3390/s25051497 ↗