Hybrid frequency-domain and machine learning framework for image noise classification and adaptive restoration
Aakanksha Jainaakanksha.jain@glsuniversity.ac.inFaculty of Computer Science EngineeringGLS UniversityAhmedabad, Gujarat, 380006, India0009-0006-2527-1773View full profile → , *Harshal ArolkarCorresponding authorharshal.arolkar@glsuniversity.ac.inFaculty of Computer Applications & Information TechnologyGLS UniversityAhmedabad, Gujarat, 380006, India0000-0003-0371-4466View full profile →
* Corresponding author · click or hover a name for details
- Received:
- 01 Apr 2025
- Published Online:
- 31 Jul 2026
- Article type:
- Research Article
- Language:
- EN
- Article no.:
- JIOS-2163
- Pages:
- 2619–2633
Abstract
Classification and restoration of envision noise are important problems in digital image processing that affect computer vision, medical imaging, and multimedia applications. This paper presents a sophisticated framework that addresses many noise types, such as Gaussian noise, impulse noise, motion blur, and speckle noise, by combining frequency-domain analysis, machine learning, and customized restoration algorithms. The method captures unique noise signatures by using the Discrete Cosine Transform (DCT), Discrete Fourier Transform (DFT) and the Fast Fourier Transform (FFT) to extract subtle frequency-domain properties. Several classifiers are investigated to achieve accurate noise type detection. A thorough assessment of the framework’s flexibility and scalability is provided by this dataset, which includes a variety of distortions and noise levels.
Adaptive, noise-specific filtering algorithms that are in line with the recognized noise type are used by the system for noise restoration. A combination of bilateral and Gaussian filters is used to handle Gaussian noise, and sophisticated median and outlier-suppression techniques are used to recover impulse noise. Adaptive and wavelet-based filters reduce speckle noise, whereas Wiener filtering and deconvolution are used to fix motion blur.
Superior performance is demonstrated by extensive experimental data, which show notable gains in classification accuracy, structural similarity index measure (SSIM), and peak signal-to-noise ratio (PSNR). By establishing a comprehensive pipeline for reliable noise classification and restoration, this study shows how machine learning and frequency-domain analysis can be used to address practical issues in image quality improvement.
Keywords
Subject Classifications
References
[1] P. Y. Liu and E. Y. Lam, “Image reconstruction using deep learning,” arXiv preprint arXiv:1809.10410 (2018).
[2] V. Jain and H. S. Seung, “Natural image denoising with convolutional networks,” in Advances in Neural Information Processing Systems, vol. 21, pp. 769–776 (2008).
[3] P. Kumar and S. K. Agarwal, “Analysis of wavelet denoising of a colour image with different types of noises,” International Journal of Signal Processing, Image Processing and Pattern Recognition, vol. 8, no. 6, pp. 125–134 (2015).
[4] M. Momeny, A. M. Latif, M. A. Sarram, R. Sheikhpour, and Y. D. Zhang, “A noise robust convolutional neural network for image classification,” Results in Engineering, vol. 10, art. no. 100225 (2021).
[5] A. E. Ilesanmi and T. O. Ilesanmi, “Methods for image denoising using convolutional neural network: A review,” Complex & Intelligent Systems, vol. 7, no. 5, pp. 2179–2198 (2021).
[6] V. M. Kamble and K. Bhurchandi, “Noise estimation and quality assessment of Gaussian noise corrupted images,” in IOP Conference Series: Materials Science and Engineering, vol. 331, no. 1, art. no. 012019 (Mar. 2018).
[7] S. S. Roy, S. I. Hossain, M. A. H. Akhand, and K. Murase, “A robust system for noisy image classification combining denoising auto-encoder and convolutional neural network,” International Journal of Advanced Computer Science and Applications, vol. 9, no. 1, pp. 224–235 (2018).
[8] M. T. Islam, S. M. Rahman, M. O. Ahmad, and M. N. S. Swamy, “Mixed Gaussian-impulse noise reduction from images using convolutional neural network,” Signal Processing: Image Communication, vol. 68, pp. 26–41 (2018).
[9] D. Sil, A. Dutta, and A. Chandra, “Convolutional neural networks for noise classification and denoising of images,” in Proc. IEEE TENCON 2019 – 2019 IEEE Region 10 Conference (TENCON), pp. 447–451 (Oct. 2019).
[10] H. Hosseini, F. Hessar, and F. Marvasti, “Real-time impulse noise suppression from images using an efficient weighted-average filtering,” IEEE Signal Processing Letters, vol. 22, no. 8, pp. 1050–1054 (2015).
[11] P. Vasuki, C. Bhavana, S. M. M. Roomi, and E. L. Deebikaa, “Automatic noise identification in images using moments and neural network,” in Proc. Int. Conf. Machine Vision and Image Processing (MVIP), pp. 61–64 (Dec. 2012).
[12] B. A. Hassan and A. A. A. Abdullah, “Enriched formulas to conjugate gradient method for removing impulse noise images,” Journal of Information and Optimization Sciences, vol. 43, no. 8, pp. 2065–2074 (2022).
[13] K. Mounika and B. R. Yadav, “Ensemble deep learning technique for optimized informative query assessment for visual images,” Journal of Information and Optimization Sciences, vol. 46, no. 2, pp. 427–437 (2025).
[14] S. Saxena, Y. Singh, B. Agarwal, and R. C. Poonia, “Comparative analysis between different edge detection techniques on mammogram images using PSNR and MSE,” Journal of Information and Optimization Sciences, vol. 43, no. 2, pp. 347–356 (2022).
[15] S. Jain, N. Rajpal, and P. K. Soni, “Multiple sclerosis detection from brain MRI images using least square-SVM,” Journal of Information and Optimization Sciences, vol. 45, no. 4, pp. 969–980 (2024).
