Abstract
In the last decade, we have witnessed significant progress in image super-resolution, thanks in particular to the emergence and improvement of deep learning models, which can adapt to the complexity of tasks and improve image quality. This article presents a novel approach to image super-resolution GA-GESRGAN to enhance the quality of document images through a multi-step methodology. Initially, document images are processed using Gabor filters to extract features across various spatial frequencies. These extracted features are then utilized to train Enhanced Super-Resolution Generative Adversarial Network (ESRGAN) models. Once improved images are obtained through ESRGAN models, a Genetic Algorithm combines these results effectively. This innovative methodology highlights the synergy between traditional image processing techniques, deep learning models, and advanced optimization algorithms, ultimately leading to significant improvements in document image quality. The results demonstrate the potential of this approach.
Recommended Citation
KEZZOULA, Zakia; GACEB, Djamel; and TITOUN, Ayoub
(2024)
"GA-GESRGAN: Document Images Super Resolution using Gabor Filters, ESRGAN Models and Genetic Algorithms,"
Emirates Journal for Engineering Research: Vol. 29:
Iss.
3, Article 2.
Available at:
https://scholarworks.uaeu.ac.ae/ejer/vol29/iss3/2