Image Enhancement and De-blurring using ResNet-101
Kusuma Gunduboina, Mohammad Nayeemul Haq, Chalcheema Sasidhar, Kakani Prudhvi, Venkatramaphanikumar Sistla, Venkata Krishna Kishore Kolli · 2025
Image de-blurring is essential for improving the quality of images across diverse fields, including photography, surveillance, and autonomous systems. This paper presents a deep learning-based approach for image deblurring using ResNet101, a deep residual network. The suggested model capitalizes on the powerful feature extraction and the potential of ResNet101 to effectively restore degraded images by reducing blur and preserving fine details. ResNet101 is designed to acquire the transformation from blurred to sharp images through its deep residual connections, allowing it to capture both global and local features essential for producing sharp, high-quality images. The training process incorporates a pixel-wise loss function and perceptual loss to ensure the deblurred images’ structural integrity and high perceptual quality. Experimental results demonstrate the efficacy of the ResNet101 based approach in significantly enhancing image clarity and sharpness.