IJSDIP | International Journal of System Design & Info Processing

Dual-Stream Transformer–CNN Hybrid Architecture for Real-Time Image Enhancement Under Extreme Noise and Degradation

Paper ID:

IJSDIP_2025_VOLUME 13_ISSUE 3_57-65

Author's Name:

S Swarna Keerthi

Volume:

Issue:

Issue 3

Year:

2025

Pages:

57-65

Abstract

Extreme noise, low illumination, and structural degradation significantly challenge conventional image enhancement models, often resulting in unstable reconstructions or loss of critical visual content. To address these limitations, this paper proposes a Dual-Stream Transformer–CNN Hybrid Architecture tailored for real-time image enhancement under severe degradation. The Transformer stream captures long-range dependencies and global illumination cues, while the CNN stream models local textures, edges, and noise patterns. A cross-stream fusion module adaptively integrates global and local representations to produce structurally consistent and perceptually rich outputs. Extensive evaluations demonstrate that the proposed framework achieves superior performance across common metrics, yielding significant gains in PSNR, SSIM, NIQE, and LPIPS. Qualitative comparisons further confirm robust detail restoration, improved contrast, and enhanced visibility in extreme low-light and noisy environments. Owing to its computational efficiency and hybrid design, the model is well suited for real-time applications in surveillance, autonomous navigation, mobile imaging, and safety-critical visual systems.

Keywords

Transformer enhancement, CNN-based reconstruction, low-light image processing, noise-robust imaging, dual-stream architecture, real-time enhancement

References

[1]. K. He, J. Sun, and X. Tang, “Single image haze removal using dark channel prior,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 33, no. 12, pp. 2341–2353, Dec. 2011. [2]. E. H. Land and J. J. McCann, “Lightness and Retinex theory,” J. Opt. Soc. Am., vol. 61, no. 1, pp. 1–11, 1971. [3]. X. Mao, C. Shen, and Y.-B. Yang, “Image restoration using very deep convolutional encoder–decoder networks with symmetric skip connections,” in Proc. NIPS, 2016, pp. 2802–2810. [4]. K. Dabov, A. Foi, V. Katkovnik, and K. Egiazarian, “Image denoising by sparse 3D transform-domain collaborative filtering,” IEEE Trans. Image Process., vol. 16, no. 8, pp. 2080–2095, Aug. 2007. [5]. M. Javeed, T. Srujana, and M. Senthil Kumar, “Design of a Clock Distribution Network Using Low Power Prescaler and Fused P & S Counters,” International Journal of Innovation in Engineering Research & Management (IJIERM), vol. 6, Special Issue 6, pp. 1–8, Oct. 2019. https://ijierm.co.in/index.php/IJIERM/article/view/1510/1658 Google Scholar [6]. K. Zhang, W. Zuo, Y. Chen, D. Meng, and L. Zhang, “Beyond a Gaussian denoiser: Residual learning of deep CNN for image denoising,” IEEE Trans. Image Process., vol. 26, no. 7, pp. 3142–3155, Jul. 2017. [7]. C. Chen, Q. Chen, J. Xu, and V. Koltun, “Learning to see in the dark,” in Proc. CVPR, 2018, pp. 3291–3300. [8]. J. Yue, T. Zhang, X. Zhou, and J. Yang, “Zero-reference deep curve estimation for low-light image enhancement,” in Proc. CVPR, 2020, pp. 1777–1786. [9]. Z. Wang, A. Bovik, H. Sheikh, and E. Simoncelli, “Image quality assessment: From error visibility to structural similarity,” IEEE Trans. Image Process., vol. 13, no. 4, pp. 600–612, Apr. 2004. [10]. S. A. Pasha, M. Javeed, and B. Laxman, “Detection Disease and Automatic Medication Composition on Plant Leafs Using Structure Algorithm,” International Journal of Innovation in Engineering Research & Management (IJIERM), vol. 6, Special Issue 6, pp. 1–8, Oct. 2019. https://ijierm.co.in/index.php/IJIERM/article/view/1517 Google Scholar [11]. J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” in Proc. NeurIPS, 2020, pp. 6840–6851. [12]. Y. Song and S. Ermon, “Generative modeling by estimating gradients of the data distribution,” in Adv. Neural Inf. Process. Syst., 2019, pp. 11895–11907. [13]. Z. Liang, J. Cao, J. Guo, and S. Wang, “TransWeather: Transformer-based restoration of images degraded by multiple weather conditions,” in Proc. CVPR, 2022, pp. 2353–2363. [14]. J. Chen, A. Li, Q. Li, J. Guo, and L. Zhang, “HSCNN: A hybrid-scale convolutional neural network for image denoising,” IEEE Trans. Image Process., vol. 29, pp. 6585–6599, 2020. [15]. MD. Javeed and C. F. Jabeen, “Linking of Cell Tracks Using Segmentation,” International Journal of Innovative Technology and Exploring Engineering (IJITEE), vol. 9, no. 1, pp. 3020–3026, Nov. 2019. https://www.ijitee.org/portfolio-item/A9128119119/ Google Scholar [16]. H. Zhao, X. Jia, and L. Sun, “SwinIR: Image restoration using Swin Transformer,” in Proc. ICCV Workshops, 2021, pp. 1833–1844. [17]. D. Liu, B. Wen, X. Liu, and W. Zeng, “Deep learning for image denoising: A survey,” IEEE Access, vol. 7, pp. 13792–13802, 2019. [18]. MD. Javeed and F. Jabeen, “FPGA Implementation for the Linking of Cell Tracks Using New Structure Algorithm,” International Journal of Engineering and Advanced Technology (IJEAT), vol. 9, no. 1, pp. 6773–6778, Oct. 2019. https://www.ijeat.org/portfolio-item/a2976109119/ Google Scholar [19]. S. Gu, L. Zhang, W. Zuo, and X. Feng, “Weighted nuclear norm minimization with application to image denoising,” in Proc. CVPR, 2014, pp. 2862–2869. [20]. J. Zamir, A. Arora, K. Khan, et al., “Restormer: Efficient Transformer for high-resolution image restoration,” in Proc. CVPR, 2022, pp. 5728–5739. [21]. M. Javeed and M. Senthil Kumar, “Automatic Brain Tumour Detection Using New Structure Algorithm,” International Journal of Innovative Technology and Exploring Engineering (IJITEE), vol. 8, no. 11, pp. 3402–3405, Sep. 2019. https://www.ijitee.org/portfolio-item/K25360981119/ Google Scholar

DOI: https://doi.org/10.64971/j.cph.ijsdip.v13.i3.8.2025