Abstract: The project targets flaws including blur, haze, scratches, color fading, and absence of color in an effort to recover old and damaged photos using a deep learning paradigm. The three GAN frameworks are integrated in a certain order to enable complicated regeneration. After patching or restoring scratches, a partial image is restored using an inpainting technique based on OpenCV. By using effective deep learning techniques, the ultimate objective is to improve the quality and accessibility of restored photos. By using deep learning and cutting-edge approaches to solve issues including blur, haze, scratches, color fading, and lack of color, the project seeks to restore old and damaged images.
Three separate GAN frameworks, each with a unique function in the restoration process, are sequentially integrated to enable complicated regeneration. After scratch patching, a OpenCV-based inpainting method is used to fill in the gaps in the image and restore a portion of it. Furthermore, certain GAN frameworks are used to manage the rest of the restoration process, making use of their individual advantages in image creation and enhancement. In the meantime, thorough restoration is ensured by the efficient detection and identification of scratches. The initiative hopes to increase the quality of recovered pictures and make them more accessible for a greater variety of uses by utilizing these advanced deep learning techniques.
Keywords: Generative Adversarial Networks (GAN), Artificial intelligence, Deep learning, OpenCV, Convolutional Neural Networks (CNNs).
Cite:
Dr. Dinesha L, Harsh Shetty, Mandira Hegde, Nesara G S, Anusha, "Regeneration Of Scratched Images Using Deep Learning", IJARCCE International Journal of Advanced Research in Computer and Communication Engineering, vol. 13, no. 3, 2024, Crossref https://doi.org/10.17148/IJARCCE.2024.13353.
|
DOI:
10.17148/IJARCCE.2024.13353
