In order to improve the denoising effect of common image denoising algorithms such as nonlocal mean image denoising, this study improved the similarity calculation formula of image blocks in the algorithm, incorporated the evaluation calculation steps of image block direction similarity and geometric information similarity, and neural network structure, and formed an improved non local mean image denoising algorithm. To test the image denoising performance of the improved algorithm, an experiment was designed and conducted in this study. The experimental data shows that the improved non local mean image denoising algorithm outperforms other comparative algorithms in terms of information entropy, peak signal-to-noise ratio, and average gradient, which represent the denoising effect. When the noise variance is 25, the normalized information entropy, peak signal-to-noise ratio, and average gradient values of the improved algorithm are 0.58, 0.62, and 0.43, respectively. The improved nonlocal mean image denoising algorithm also outperforms other algorithms in the evaluation index of structure similarity, mutual information amount, root mean square deviation, and absolute error over the original information retention. When the noise variance is 45, the normalized structure similarity, Mutual information amount, root mean square deviation, and absolute error of the algorithm are 0.88, 0.78, 8.2, and 9.7, respectively. However, the computational efficiency of the improved non local mean image denoising algorithm is slightly slower than that ofthe traditional non local mean image denoising algorithms. It can be seen that the improved algorithm designed in this study has a superior denoising performance compared to common methods, and can retain the core information of the original image. However, its computational efficiency is slow, and it can provide better image denoising services for applications with low computational efficiency requirements, such as meteorological prediction, agricultural planting, and other fields.