This paper presents a comprehensive approach to image denoising and color restoration using artificial intelligence techniques. We investigate how convolutional neural networks (CNNs) can be leveraged to enhance image quality by removing noise artifacts and restoring color to grayscale images. Our methodology employs a dual-model approach: one model dedicated to noise reduction and another focused on color restoration. Both models utilize residual connections and specialized normalization techniques to preserve image details while performing their respective tasks. Experiments conducted on the CIFAR-10 and ImageNet datasets demonstrate the effectiveness of our approach, with quantitative evaluations using peak signal-to-noise ratio (PSNR) and qualitative visual assessments. The denoising model successfully removes Gaussian and salt-and-pepper noise while preserving essential image features, achieving a PSNR of up to 22.78~dB. The color restoration model transforms grayscale images into plausible color representations, with results improving significantly over extended training periods. This research contributes to the field of image processing by providing insights into neural network architectures optimized for image enhancement tasks and demonstrating their practical applications in restoring degraded visual content. Additionally, this work supports UN Sustainable Development Goal~9 (Industry, Innovation and Infrastructure) by advancing AI-based techniques for restoring degraded images, thereby enhancing the reliability and efficiency of digital imaging systems in modern industrial and technological applications.
 

Image denoising, color restoration, image colorization, convolutional neural networks (CNNs), residual learning, normalization, Gaussian noise, salt-and-pepper noise, CIFAR-10, ImageNet, peak signal-to-noise ratio (PSNR), image enhancement, SDG 9.