"Hyperspectral Image (HSI) is a three-dimensional image data captured by spectrometers. However, during hyperspectral imaging, it often encounters interference and contamination from various types of noise, which hinders the subsequent analysis and application of HSIs, making denoising of high importance. Although the application of Transformers has effectively improved the performance of HSI denoising, the original Transformer model still has limitations in utilizing the spatial-spectral coupling correlation within HSIs. It tends to over-smooth spatial features, leading to the loss of small-scale structures, and overly focuses on all channel features in the spectral dimension, lacking the utilization of differences among different spectral bands. To address these issues, we propose a novel sparse spatial-spectral Transformer model. In the spatial dimension, we introduce a local convolution module to enhance spatial feature details and mitigate the over-smoothing issue. Simultaneously, in the spectral dimension, we propose a Top-k sparse self-attention mechanism to adaptively select the most relevant spectral channel information for feature interactions, effectively capturing spatial-spectral features. HSI denoising is achieved through hierarchical residual connections in the sparse spatial-spectral Transformer. Experimental results demonstrate that our proposed method outperforms current state-of-the-art approaches under Gaussian noise and complex mixed noise conditions. Furthermore, ablation experiments validate the effectiveness of our sparse spatial-spectral Transformer model."