Research on Lightweight Super-Resolution Reconstruction Algorithm Based on Progressive Feature Fusion Convolutional

This paper proposes a lightweight super-resolution reconstruction algorithm based on a progressive feature fusion convolutional network. The algorithm improves image resolution by incorporating a progressive attention block and a multi-scale pixel attention mechanism to reduce computational complexity. The progressive attention block uses multiple parallel paths to gradually aggregate multi-scale features, achieving a larger receptive field with fewer parameters. The multi-scale pixel attention mechanism builds an efficient upsampling module, ensuring effective utilization of the attention mechanism while maintaining lightweight characteristics. Additionally, the paper introduces a cosine annealing-based training strategy to enhance model expressiveness and image restoration quality without altering the model structure. The experimental results show that the algorithm proposed in this paper performs better than current mainstream algorithms in terms of running speed and efficiency balance.