An On-Device AI Based Video Compression Model Optimization for Efficient Bandwidth Usage and Image Quality in Constrained AIoT Environments

The rapid expansion of video data from applications such as surveillance, streaming, and real-time systems has amplified the need for advanced compression techniques, especially in bandwidth-limited AIoT environments. Conventional codecs like H.264 and H.265 often struggle to balance compression rates with image quality, particularly in scenes with high levels of motion. This paper introduces an AI-driven video compression model designed to optimize transmission efficiency by utilizing a background image repository. By isolating dynamic foreground elements and transmitting only the relevant data, the model substantially reduces bandwidth consumption without compromising visual quality. Experimental results show that the proposed method achieves an effective balance between compression efficiency and image fidelity, consistently outperforming traditional codecs in bandwidth usage and visual quality trade-offs. This model is particularly well-suited for real-time video transmission in resource-constrained AIoT settings, such as surveillance networks and remote monitoring systems.