The Information Multi - Distillation Network (IMDN) represents an efficient image super - resolution model. To further optimize the lightweight performance of IMDN, this study presents two independent improvement strategies: Asymmetric Channel Splitting (AMCS) and Nested Multi - Information Distillation Block (NIMDB). AMCS divides the input features into two branches with unequal numbers of channels, thereby enhancing the super - resolution reconstruction capabilities for different features. Specifically, this approach enables the network to better handle and reconstruct various feature components. NIMDB, on the other hand, replaces the convolutional layers within the distillation block with another distillation block, establishing a nested distillation architecture. This structure allows for a more refined extraction of features while maintaining the original network framework. Experimental results on the lightweight model IMDN - RTC and the original IMDN demonstrate that AMCS can significantly enhance the performance of lightweight models. Specifically, compared to the lightweight IMDN - RTC, the Peak Signal - to - Noise Ratio (PSNR) is improved by up to 0.16. NIMDB, conversely, is more compatible with the original model structure. In comparison to the original IMDN, it achieves a maximum PSNR improvement of 0.09 and a maximum improvement of 0.0014 in the Structural Similarity Index (SSIM). By adapting suitable lightweight strategies to models of different scales, this research not only reduces the parameter count by 5% and 11% respectively but also maximizes the improvement of both PSNR and SSIM.