Deep learning-based adaptive image compression and narrowband communication bandwidth matching optimization system

The growing requirement to deliver high-resolution pictures over NB channels poses a significant problem: maintaining picture quality while minimizing bandwidth. Common compression algorithms typically employ constant bit rate encoding, which fails to adapt to changing network conditions, resulting in either inferior image quality or excessive transmission latency. Hence, a method that mediates between compression effectiveness and flexible bandwidth control is required. The suggested method presents an adaptive Image Compression (IC) and bandwidth matching method embedding a Rate-Distortion Auto-encoder enhanced Shrike Algorithm (RDAES). High-res images are preprocessed with resizing, MM normalization, and RGB Tensor Preparation for consistency in DL model input. A Convolutional Neural Network (CNN) captures multi-level spatial features and is followed by a Rate-Distortion Optimized Auto-encoder, obtaining concise latent features. Shrike Optimization Algorithm (SOA) compresses images with bandwidth capacity, increasing efficiency and fidelity. RDAE dynamically encodes images with reduced artefacts and great detail. Evaluation is conducted on accessible datasets with various natural and surveillance images, guaranteeing generalization capabilities. RDAES model achieves peak PSNR (49.90 dB), MS-SSIM (0.90), and LPIPS ( 0.23). Implementation leverages Python, with adaptive modules validated under simulated NB channel conditions. Results demonstrate substantial improvements in bandwidth utilization and image reconstruction quality compared to fixed-rate or conventional compression methods. The RDAES model provides a scalable, adaptive, and efficient solution for high-fidelity image transmission over constrained communication channels, ensuring low latency and high reliability.