Enhancing Single-Image Super-Resolution using Patch-Mosaic Data Augmentation on Lightweight Bimodal Network

With the advancement of deep learning, single-image super-resolution (SISR) has made significant strides. However, most current SISR methods are challenging to employ in real-world applications because they doubtlessly employed by substantial computational and memory costs caused complex operations. Furthermore, an efficient dataset is a key factor for bettering model training. The hybrid models CNN Vision Transformer can be more task. Nevertheless, require or extremely high-quality datasets training that could unavailable from time time. To tackle these issues, solution combined applying Lightweight Bimodal Network (LBNet) Patch-Mosaic data augmentation method which enhancement CutMix YOCO proposed this research. patch-oriented Mosaic augmentation, Symmetric utilized local feature extraction coarse image restoration. Plus, Recursive aids fully grasping long-term dependence images, enabling global information used refine texture details. Extensive experiments have shown LBNet with zero-free additional parameters outperforms original other state-of-the-art techniques image-level applied.