THE ROLE OF MODERN ARTIFICIAL INTELLIGENCE IN DIGITAL IMAGE PROCESSING: A COMPREHENSIVE LITERATURE REVIEW (2021-2026)

Abstract

This review summarizes the recent advances of digital image processing using artificial intelligence methods from 2021 to 2026. It explores the development, performance and applications of convolutional neural networks, Vision Transformers, generative adversarial networks, U-Net variants and hybrid architectures as well as their limitations and future directions. Direct comparison of a few studies that were reviewed to the others is limited with differences in data sets, validation methods, class distributions, and evaluation methods. Several studies reviewed reported accuracy greater than 98% on specific standard data sets, but direct comparison is limited since the data sets, validation methods, class distributions and evaluation methods are different. In addition to healthcare, AI capabilities have been proven to be quite impressive with respect to autonomous driving, remote sensing, industrial quality control, and computer vision. New innovations include the integration of several AI paradigms, attention mechanisms for improved feature extraction, and lightweight models for resource-constrained applications. While there have been great strides, there are still obstacles to overcome in terms of model interpretability, data harmonization, computational efficiency and clinical deployment. It is a review that presents an overview of existing methodologies, comparative performance measures, and emerging trends, offering researchers and practitioners a solid foundation at the intersection of AI and digital image processing.