A survey of edge AI techniques for real-time image processing in intelligent visual computing systems

Recent developments in AI, IoT, and computer vision have driven the adoption of intelligent image processing systems capable of handling visual data with greater speed and accuracy. However, traditional cloud-based solutions often struggle in real-time applications due to latency issues, high bandwidth consumption, and growing concerns related to data privacy. These limitations make them less practical for time-critical and resource-sensitive environments.
Edge Artificial Intelligence (Edge AI) has emerged as a strong alternative by shifting computation from centralized cloud servers to edge devices where data is generated. This approach reduces communication delay, minimizes network dependency, and improves system responsiveness in real-world scenarios. This survey presents a structured overview of Edge AI techniques used for real-time image processing in intelligent visual computing systems. Particular attention is given to compact deep learning models and optimization approaches tailored for edge platforms with limited processing and memory capabilities.
The paper also discusses key application areas such as healthcare monitoring, intelligent transportation, precision agriculture, and industrial automation. In addition, it highlights major challenges including energy efficiency, hardware limitations, and secure model deployment. Finally, emerging research directions are outlined, particularly those aimed at improving efficiency, scalability, and real-world adaptability of Edge AI systems.