Image Regeneration for Old Damaged Real Picture
Aswathy V
, Irene Jerusha P , Nuvvuru Divya
Generative adversarial network, Resnet-50, Parsenet, Convolutional Neural Network, Image Preprocessing
Old photos are little things that keep memories alive. Each photograph is worth a thousand memories. However, several photographs have permanent deterioration as a result of unsuitable environmental conditions that lead to photodegradation. Motion blur, noise, and camera focus issues are some forms of degradation. Photo restoration is a technology to repair old photos for those who wish to bring old photos back to life. The two primary techniques utilized in conventional picture restoration to replace the destroyed areas of the image are thermal diffusion and mathematical calculations. It is challenging for this technology to be incorporated into people's daily lives since it can only restore photographs with simple structures and minimal deterioration. Nonetheless, there are still many of improvements being made in this time-honored method of picture restoration. A triplet domain translation network may now be used to stop the combined degradation of old photographs. When the translation to clean images is learned in latent space, the domain gap between historical photos and synthetic images is reduced. When compared to previous methods, this has a generalization problem, which is its main disadvantage. The development of deep learning technologies has boosted the scope of image restoration research. This project's primary goal is to offer an image restoration method based on the Resnet-50, GFP-GAN, and Parsenet algorithms in order to enhance the results of picture restoration of old images and provide additional possibilities for the procedure. The structure, concept, and loss function of an image restoration model built on Resnet-50, GFP-GAN, and Parsenet are shown. In order to create deblurred images and repair old, damaged photographs, GFP-GAN and Parsenet are employed, while Resnet-50 is used to train exceptionally deep neural networks. The historical significance of image restoration techniques is also discussed in this paper. The project concluded that, in the experiment for blur correction, its algorithm was superior to other algorithms. This was done by conducting a comparative experiment to compare this model to other models. The study's model, then, has the greatest effect on picture restoration.
"Image Regeneration for Old Damaged Real Picture", TIJER - TIJER - INTERNATIONAL RESEARCH JOURNAL (www.TIJER.org), ISSN:2349-9249, Vol.11, Issue 2, page no.105-110, February-2024, Available :https://tijer.org/TIJER/papers/TIJERB001016.pdf
Volume 11
Issue 2,
February-2024
Pages : 105-110
Paper Reg. ID: TIJER_150944
Published Paper Id: TIJERB001016
Downloads: 000397
Research Area: Science and Technology
Country: Thiruvallur, TamilNadu, India
ISSN: 2349-9249 | IMPACT FACTOR: 8.57 Calculated By Google Scholar | ESTD YEAR: 2014
An International Scholarly Open Access Journal, Peer-Reviewed, Refereed Journal Impact Factor 8.57 Calculate by Google Scholar and Semantic Scholar | AI-Powered Research Tool, Multidisciplinary, Monthly, Multilanguage Journal Indexing in All Major Database & Metadata, Citation Generator
Publisher: TIJER (IJ Publication) Janvi Wave
