A Recursive Pooling with a Metaheuristic Aware Video Compression and Super Resolution Model Based on Deep Learning Technique

The rapid growth of high-quality and immersive video content has posed substantial problems to compression methods, as traditional approaches frequently fail to reconcile bitrate reduction with perceived quality. Many existing techniques have apparent artifacts, limited generalization across various datasets, and poor performance in super-resolution tasks, limiting their usefulness in real-world applications. To address these limitations, this study presents a deep learning-based dual-compression system built on the ResNet152 architecture. The suggested model has used an innovative compression strategy using the ResNet152 architecture. The suggested model incorporates the Weighted Learning Discrete Wavelet Pooling (WLD-WP) and Recurrent Gradient Filter (RGF) into its design. The Multi-Layer Perception offers advanced categorization to a high degree. The dual-compressive framework addresses the reduction of compression artifacts and the enhancement of super-resolution. It utilizes a meta-heuristic approach that incorporates the Horse Herd Optimization (HHO) algorithm and the Gannet Optimization Algorithm (GOA). This hybrid approach is referred to as Gannet-Enriched HHO (GE-HHO). The results are evaluated by implementing in a MATLAB platform, in which the suggested model achieved a better performance and lower error rates. Hence, the outcomes demonstrate that the suggested model outperforms the other models.