Deep learning (DL) has revolutionized image processing by enabling performance far beyond traditional techniques. This survey reviews the evolution of DL-based image processing methods, from early architectures to state-of-the-art models and learning paradigms. It highlights key advancements that enhance efficiency, generalization, and robustness for analyzing complex visual data across diverse applications. Commonly used evaluation metrics are discussed to emphasize rigorous performance assessment. The survey also outlines future research directions, including quantum and neuromorphic computing, federated learning for privacy-preserving training, and the integration of edge computing and explainable artificial intelligence to address scalability and interpretability challenges.
Babu et al. (Thu,) studied this question.