Research on End-to-End Image Dehazing Algorithm Based on Ultra-Lightweight Network Architectures for Efficient Visual Enhancement

It is a common problem in computer vision applications in which the visibility of images is reduced due to haze, and lots of detail is lost. To solve this problem, we propose an end-to-end single-image de hazing algorithm based on ultra-lightweight network architecture, which combines the existing advance methods and innovates based on this to achieve efficient visual enhancement. Inspired from the recent advances in existing deep learning-based dehazing models, we introduce a multi-scale feature extraction module which helps capture local and global cues related to haze with computational efficiency. The integration of residual learning strategy and attention mechanism refines the model in the process of dehazing and improves the accuracy of dehazing with the lightweight and high efficiency retained. Furthermore, the hybrid loss function is optimized to achieve a harmony between pixel-level accuracy and perceptual quality. The experimental results indicate that the proposed model outperforms the existing dehazing models in terms of visual quality and computational efficiency.