Enhancing Image
Clarity: Advanced Deep Learning Strategies for Noise Reduction and Image Restoration
for Disaster Management
Chopparapu SaiTeja, Chopparapu Gowthami, Mamidi Ranjeeth Reddy, Badithala Sravan
Kumar and Beera Jaya Bharathi
Disaster Advances; Vol. 19(1); 17-28;
doi: https://doi.org/10.25303/191da017028; (2026)
Abstract
Image denoising is a crucial task in digital image processing, but recent State-of-the-Art
models, while powerful, are often too large and slow for practical applications.
This creates a significant gap between theoretical performance and real-world deployability.
This research introduces a novel deep learning framework that bridges this gap.
The key novelty lies in the synergistic integration of a Dual Path Network (DPN)
for efficient feature extraction, a Convolutional Block Attention Module (CBAM)
to preserve fine details and a Generative Adversarial Network (GAN) for perceptual
realism. Evaluated on standard benchmarks, the proposed model achieves a Peak Signal-to-Noise
Ratio (PSNR) up to 34.02 dB and a Structural Similarity Index Measure (SSIM) up
to 0.932, consistently outperforming traditional methods and recent State-of-the-Art
architectures from 2024-2025. Crucially, the model's primary advantage is its exceptional
efficiency.
With only 0.55 million parameters and an inference time of 35 ms, it is over 80
times smaller and twice as fast as leading competitors, without sacrificing performance.
This outstanding balance of accuracy and efficiency makes the proposed framework
a practical and superior solution for high-quality, real-time denoising on resource-constrained
devices such as in disaster management and embedded vision systems.
