The paper introduces a heterogeneous window Transformer (HWformer) for image denoising, aiming to improve denoising performance by capturing both global and local context information. HWformer addresses the limitations of traditional deep networks, which often rely on short-distance modeling and may overlook pixel correlations. To achieve this, HWformer employs heterogeneous global windows to capture rich global context, and shifts these windows horizontally and vertically to facilitate diverse interactions without increasing denoising time. Additionally, a sparse technique is applied to a feed-forward network to extract local information from neighboring patches, preventing information loss. The proposed HWformer is designed to balance denoising speed and effectiveness, achieving 30% of the denoising time compared to popular Restormer while maintaining or improving denoising quality. The paper includes experimental results on various datasets, demonstrating HWformer's superior performance in terms of PSNR, SSIM, FSIM, LPIPS, PSBR, and color difference, making it suitable for real-world applications such as smartphones and cameras.The paper introduces a heterogeneous window Transformer (HWformer) for image denoising, aiming to improve denoising performance by capturing both global and local context information. HWformer addresses the limitations of traditional deep networks, which often rely on short-distance modeling and may overlook pixel correlations. To achieve this, HWformer employs heterogeneous global windows to capture rich global context, and shifts these windows horizontally and vertically to facilitate diverse interactions without increasing denoising time. Additionally, a sparse technique is applied to a feed-forward network to extract local information from neighboring patches, preventing information loss. The proposed HWformer is designed to balance denoising speed and effectiveness, achieving 30% of the denoising time compared to popular Restormer while maintaining or improving denoising quality. The paper includes experimental results on various datasets, demonstrating HWformer's superior performance in terms of PSNR, SSIM, FSIM, LPIPS, PSBR, and color difference, making it suitable for real-world applications such as smartphones and cameras.